id
stringlengths 9
16
| submitter
stringlengths 3
64
⌀ | authors
stringlengths 5
6.63k
| title
stringlengths 7
245
| comments
stringlengths 1
482
⌀ | journal-ref
stringlengths 4
382
⌀ | doi
stringlengths 9
151
⌀ | report-no
stringclasses 984
values | categories
stringlengths 5
108
| license
stringclasses 9
values | abstract
stringlengths 83
3.41k
| versions
listlengths 1
20
| update_date
timestamp[s]date 2007-05-23 00:00:00
2025-04-11 00:00:00
| authors_parsed
sequencelengths 1
427
| prompt
stringlengths 166
3.49k
| label
stringclasses 2
values | prob
float64 0.5
0.98
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1104.3538 | Henrique Araujo | H. M. Araujo, D. Yu. Akimov, E. J. Barnes, V. A. Belov, A. Bewick, A.
A. Burenkov, V. Chepel. A. Currie, L. DeViveiros, B. Edwards, C. Ghag, A.
Hollingsworth, M. Horn, G. E. Kalmus, A. S. Kobyakin, A. G. Kovalenko, V. N.
Lebedenko, A. Lindote, M. I. Lopes, R. Luscher, P. Majewski, A. StJ. Murphy.
F. Neves, S. M. Paling, J. Pinto da Cunha, R. Preece, J. J. Quenby, L.
Reichhart, P. R. Scovell, C. Silva, V. N. Solovov, N. J. T. Smith, P. F.
Smith, V. N. Stekhanov, T. J. Sumner, C. Thorne, R. J. Walker | Radioactivity Backgrounds in ZEPLIN-III | 12 pages, 5 figures | null | 10.1016/j.astropartphys.2011.11.001 | null | physics.ins-det hep-ex | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We examine electron and nuclear recoil backgrounds from radioactivity in the
ZEPLIN-III dark matter experiment at Boulby. The rate of low-energy electron
recoils in the liquid xenon WIMP target is 0.75$\pm$0.05 events/kg/day/keV,
which represents a 20-fold improvement over the rate observed during the first
science run. Energy and spatial distributions agree with those predicted by
component-level Monte Carlo simulations propagating the effects of the
radiological contamination measured for materials employed in the experiment.
Neutron elastic scattering is predicted to yield 3.05$\pm$0.5 nuclear recoils
with energy 5-50 keV per year, which translates to an expectation of 0.4 events
in a 1-year dataset in anti-coincidence with the veto detector for realistic
signal acceptance. Less obvious background sources are discussed, especially in
the context of future experiments. These include contamination of scintillation
pulses with Cherenkov light from Compton electrons and from $\beta$ activity
internal to photomultipliers, which can increase the size and lower the
apparent time constant of the scintillation response. Another challenge is
posed by multiple-scatter $\gamma$-rays with one or more vertices in regions
that yield no ionisation. If the discrimination power achieved in the first run
can be replicated, ZEPLIN-III should reach a sensitivity of $\sim 1 \times
10^{-8}$ pb$\cdot$year to the scalar WIMP-nucleon elastic cross-section, as
originally conceived.
| [
{
"version": "v1",
"created": "Mon, 18 Apr 2011 16:45:26 GMT"
},
{
"version": "v2",
"created": "Fri, 12 Aug 2011 13:19:49 GMT"
}
] | 2015-05-27T00:00:00 | [
[
"Araujo",
"H. M.",
""
],
[
"Akimov",
"D. Yu.",
""
],
[
"Barnes",
"E. J.",
""
],
[
"Belov",
"V. A.",
""
],
[
"Bewick",
"A.",
""
],
[
"Burenkov",
"A. A.",
""
],
[
"Currie",
"V. Chepel. A.",
""
],
[
"DeViveiros",
"L.",
""
],
[
"Edwards",
"B.",
""
],
[
"Ghag",
"C.",
""
],
[
"Hollingsworth",
"A.",
""
],
[
"Horn",
"M.",
""
],
[
"Kalmus",
"G. E.",
""
],
[
"Kobyakin",
"A. S.",
""
],
[
"Kovalenko",
"A. G.",
""
],
[
"Lebedenko",
"V. N.",
""
],
[
"Lindote",
"A.",
""
],
[
"Lopes",
"M. I.",
""
],
[
"Luscher",
"R.",
""
],
[
"Majewski",
"P.",
""
],
[
"Neves",
"A. StJ. Murphy. F.",
""
],
[
"Paling",
"S. M.",
""
],
[
"da Cunha",
"J. Pinto",
""
],
[
"Preece",
"R.",
""
],
[
"Quenby",
"J. J.",
""
],
[
"Reichhart",
"L.",
""
],
[
"Scovell",
"P. R.",
""
],
[
"Silva",
"C.",
""
],
[
"Solovov",
"V. N.",
""
],
[
"Smith",
"N. J. T.",
""
],
[
"Smith",
"P. F.",
""
],
[
"Stekhanov",
"V. N.",
""
],
[
"Sumner",
"T. J.",
""
],
[
"Thorne",
"C.",
""
],
[
"Walker",
"R. J.",
""
]
] | TITLE: Radioactivity Backgrounds in ZEPLIN-III
ABSTRACT: We examine electron and nuclear recoil backgrounds from radioactivity in the
ZEPLIN-III dark matter experiment at Boulby. The rate of low-energy electron
recoils in the liquid xenon WIMP target is 0.75$\pm$0.05 events/kg/day/keV,
which represents a 20-fold improvement over the rate observed during the first
science run. Energy and spatial distributions agree with those predicted by
component-level Monte Carlo simulations propagating the effects of the
radiological contamination measured for materials employed in the experiment.
Neutron elastic scattering is predicted to yield 3.05$\pm$0.5 nuclear recoils
with energy 5-50 keV per year, which translates to an expectation of 0.4 events
in a 1-year dataset in anti-coincidence with the veto detector for realistic
signal acceptance. Less obvious background sources are discussed, especially in
the context of future experiments. These include contamination of scintillation
pulses with Cherenkov light from Compton electrons and from $\beta$ activity
internal to photomultipliers, which can increase the size and lower the
apparent time constant of the scintillation response. Another challenge is
posed by multiple-scatter $\gamma$-rays with one or more vertices in regions
that yield no ionisation. If the discrimination power achieved in the first run
can be replicated, ZEPLIN-III should reach a sensitivity of $\sim 1 \times
10^{-8}$ pb$\cdot$year to the scalar WIMP-nucleon elastic cross-section, as
originally conceived.
| no_new_dataset | 0.93276 |
1104.4512 | Vassilis Kekatos | Pedro A. Forero, Vassilis Kekatos, Georgios B. Giannakis | Robust Clustering Using Outlier-Sparsity Regularization | Submitted to IEEE Trans. on PAMI | null | 10.1109/TSP.2012.2196696 | null | stat.ML cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Notwithstanding the popularity of conventional clustering algorithms such as
K-means and probabilistic clustering, their clustering results are sensitive to
the presence of outliers in the data. Even a few outliers can compromise the
ability of these algorithms to identify meaningful hidden structures rendering
their outcome unreliable. This paper develops robust clustering algorithms that
not only aim to cluster the data, but also to identify the outliers. The novel
approaches rely on the infrequent presence of outliers in the data which
translates to sparsity in a judiciously chosen domain. Capitalizing on the
sparsity in the outlier domain, outlier-aware robust K-means and probabilistic
clustering approaches are proposed. Their novelty lies on identifying outliers
while effecting sparsity in the outlier domain through carefully chosen
regularization. A block coordinate descent approach is developed to obtain
iterative algorithms with convergence guarantees and small excess computational
complexity with respect to their non-robust counterparts. Kernelized versions
of the robust clustering algorithms are also developed to efficiently handle
high-dimensional data, identify nonlinearly separable clusters, or even cluster
objects that are not represented by vectors. Numerical tests on both synthetic
and real datasets validate the performance and applicability of the novel
algorithms.
| [
{
"version": "v1",
"created": "Fri, 22 Apr 2011 22:01:14 GMT"
}
] | 2015-05-27T00:00:00 | [
[
"Forero",
"Pedro A.",
""
],
[
"Kekatos",
"Vassilis",
""
],
[
"Giannakis",
"Georgios B.",
""
]
] | TITLE: Robust Clustering Using Outlier-Sparsity Regularization
ABSTRACT: Notwithstanding the popularity of conventional clustering algorithms such as
K-means and probabilistic clustering, their clustering results are sensitive to
the presence of outliers in the data. Even a few outliers can compromise the
ability of these algorithms to identify meaningful hidden structures rendering
their outcome unreliable. This paper develops robust clustering algorithms that
not only aim to cluster the data, but also to identify the outliers. The novel
approaches rely on the infrequent presence of outliers in the data which
translates to sparsity in a judiciously chosen domain. Capitalizing on the
sparsity in the outlier domain, outlier-aware robust K-means and probabilistic
clustering approaches are proposed. Their novelty lies on identifying outliers
while effecting sparsity in the outlier domain through carefully chosen
regularization. A block coordinate descent approach is developed to obtain
iterative algorithms with convergence guarantees and small excess computational
complexity with respect to their non-robust counterparts. Kernelized versions
of the robust clustering algorithms are also developed to efficiently handle
high-dimensional data, identify nonlinearly separable clusters, or even cluster
objects that are not represented by vectors. Numerical tests on both synthetic
and real datasets validate the performance and applicability of the novel
algorithms.
| no_new_dataset | 0.951729 |
1406.5141 | Adrian Melott | A.L. Melott (Kansas), I.G. Usoskin (Oulu), G.A Kovaltsov (St.
Petersburg), and C.M. Laird (Kansas) | Has the Earth been exposed to numerous supernovae within the last 300
kyr? | 8 pages, 1 figure. to be published in the Interenational Journal of
Astrobiology | International Journal of Astrobiology 14, 375-378 (2015) | 10.1017/S1473550414000512 | null | astro-ph.EP astro-ph.HE astro-ph.SR physics.geo-ph q-bio.PE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Firestone (2014) asserted evidence for numerous (23) nearby (d<300 pc)
supernovae within the Middle and Late Pleistocene. If true, this would have
strong implications for the irradiation of the Earth; at this rate, mass
extinction level events due to supernovae would be more frequent than 100 Myr.
However, there are numerous errors in the application of past research. The
paper overestimates likely nitrate and 14C production from moderately nearby
supernovae by about four orders of magnitude. Moreover, the results are based
on wrongly selected (obsolete) nitrate and 14C datasets. The use of correct and
up-to-date datasets does not confirm the claimed results. The claims in the
paper are invalidated.
| [
{
"version": "v1",
"created": "Thu, 19 Jun 2014 18:36:33 GMT"
},
{
"version": "v2",
"created": "Thu, 25 Sep 2014 14:21:05 GMT"
}
] | 2015-05-27T00:00:00 | [
[
"Melott",
"A. L.",
"",
"Kansas"
],
[
"Usoskin",
"I. G.",
"",
"Oulu"
],
[
"Kovaltsov",
"G. A",
"",
"St.\n Petersburg"
],
[
"Laird",
"C. M.",
"",
"Kansas"
]
] | TITLE: Has the Earth been exposed to numerous supernovae within the last 300
kyr?
ABSTRACT: Firestone (2014) asserted evidence for numerous (23) nearby (d<300 pc)
supernovae within the Middle and Late Pleistocene. If true, this would have
strong implications for the irradiation of the Earth; at this rate, mass
extinction level events due to supernovae would be more frequent than 100 Myr.
However, there are numerous errors in the application of past research. The
paper overestimates likely nitrate and 14C production from moderately nearby
supernovae by about four orders of magnitude. Moreover, the results are based
on wrongly selected (obsolete) nitrate and 14C datasets. The use of correct and
up-to-date datasets does not confirm the claimed results. The claims in the
paper are invalidated.
| no_new_dataset | 0.942612 |
1505.06751 | Ayad Ghany Ismaeel | Ayad Ghany Ismaeel, Raghad Zuhair Yousif | Novel Mining of Cancer via Mutation in Tumor Protein P53 using Quick
Propagation Network | 6 Pages, 9 figures, 2 Table | International Journal of Computer Science and Electronics
Engineering IJCSEE, Volume 3, Issue 2, 2015 | null | null | cs.CE | http://creativecommons.org/licenses/by/3.0/ | There is multiple databases contain datasets of TP53 gene and its tumor
protein P53 which believed to be involved in over 50% of human cancers cases,
these databases are rich as datasets covered all mutations caused diseases
(cancers), but they haven't efficient mining method can classify and diagnosis
mutations patient's then predict the cancer of that patient. This paper
proposed a novel mining of cancer via mutations because there is no mining
method before offers friendly, effective and flexible predict or diagnosis of
cancers via using whole common database of TP53 gene (tumor protein P53) as
dataset and selecting a minimum number of fields in training and testing quick
propagation algorithm which supporting this miming method. Simulating quick
propagation network for the train dataset shows results the Correlation
(0.9999), R-squared (0.9998) and mean of Absolute Relative Error (0.0029),
while the training for the ALL datasets (train, test and validation dataset)
have results the Correlation (0.9993), R-squared (0.9987) and mean of Absolute
Relative Error (0.0057).
| [
{
"version": "v1",
"created": "Wed, 20 May 2015 15:46:45 GMT"
}
] | 2015-05-27T00:00:00 | [
[
"Ismaeel",
"Ayad Ghany",
""
],
[
"Yousif",
"Raghad Zuhair",
""
]
] | TITLE: Novel Mining of Cancer via Mutation in Tumor Protein P53 using Quick
Propagation Network
ABSTRACT: There is multiple databases contain datasets of TP53 gene and its tumor
protein P53 which believed to be involved in over 50% of human cancers cases,
these databases are rich as datasets covered all mutations caused diseases
(cancers), but they haven't efficient mining method can classify and diagnosis
mutations patient's then predict the cancer of that patient. This paper
proposed a novel mining of cancer via mutations because there is no mining
method before offers friendly, effective and flexible predict or diagnosis of
cancers via using whole common database of TP53 gene (tumor protein P53) as
dataset and selecting a minimum number of fields in training and testing quick
propagation algorithm which supporting this miming method. Simulating quick
propagation network for the train dataset shows results the Correlation
(0.9999), R-squared (0.9998) and mean of Absolute Relative Error (0.0029),
while the training for the ALL datasets (train, test and validation dataset)
have results the Correlation (0.9993), R-squared (0.9987) and mean of Absolute
Relative Error (0.0057).
| no_new_dataset | 0.946051 |
1505.06800 | Baochang Zhang | Lei Wang, Baochang Zhang | Boosting-like Deep Learning For Pedestrian Detection | 9 pages,7 figures | null | null | null | cs.CV cs.LG cs.NE | http://creativecommons.org/licenses/by-nc-sa/3.0/ | This paper proposes boosting-like deep learning (BDL) framework for
pedestrian detection. Due to overtraining on the limited training samples,
overfitting is a major problem of deep learning. We incorporate a boosting-like
technique into deep learning to weigh the training samples, and thus prevent
overtraining in the iterative process. We theoretically give the details of
derivation of our algorithm, and report the experimental results on open data
sets showing that BDL achieves a better stable performance than the
state-of-the-arts. Our approach achieves 15.85% and 3.81% reduction in the
average miss rate compared with ACF and JointDeep on the largest Caltech
benchmark dataset, respectively.
| [
{
"version": "v1",
"created": "Tue, 26 May 2015 03:52:52 GMT"
}
] | 2015-05-27T00:00:00 | [
[
"Wang",
"Lei",
""
],
[
"Zhang",
"Baochang",
""
]
] | TITLE: Boosting-like Deep Learning For Pedestrian Detection
ABSTRACT: This paper proposes boosting-like deep learning (BDL) framework for
pedestrian detection. Due to overtraining on the limited training samples,
overfitting is a major problem of deep learning. We incorporate a boosting-like
technique into deep learning to weigh the training samples, and thus prevent
overtraining in the iterative process. We theoretically give the details of
derivation of our algorithm, and report the experimental results on open data
sets showing that BDL achieves a better stable performance than the
state-of-the-arts. Our approach achieves 15.85% and 3.81% reduction in the
average miss rate compared with ACF and JointDeep on the largest Caltech
benchmark dataset, respectively.
| no_new_dataset | 0.948917 |
1505.06812 | Purushottam Kar | Harikrishna Narasimhan and Purushottam Kar and Prateek Jain | Optimizing Non-decomposable Performance Measures: A Tale of Two Classes | To appear in proceedings of the 32nd International Conference on
Machine Learning (ICML 2015) | Journal of Machine Learning Research, W&CP 37 (2015) | null | null | stat.ML cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Modern classification problems frequently present mild to severe label
imbalance as well as specific requirements on classification characteristics,
and require optimizing performance measures that are non-decomposable over the
dataset, such as F-measure. Such measures have spurred much interest and pose
specific challenges to learning algorithms since their non-additive nature
precludes a direct application of well-studied large scale optimization methods
such as stochastic gradient descent.
In this paper we reveal that for two large families of performance measures
that can be expressed as functions of true positive/negative rates, it is
indeed possible to implement point stochastic updates. The families we consider
are concave and pseudo-linear functions of TPR, TNR which cover several
popularly used performance measures such as F-measure, G-mean and H-mean.
Our core contribution is an adaptive linearization scheme for these families,
using which we develop optimization techniques that enable truly point-based
stochastic updates. For concave performance measures we propose SPADE, a
stochastic primal dual solver; for pseudo-linear measures we propose STAMP, a
stochastic alternate maximization procedure. Both methods have crisp
convergence guarantees, demonstrate significant speedups over existing methods
- often by an order of magnitude or more, and give similar or more accurate
predictions on test data.
| [
{
"version": "v1",
"created": "Tue, 26 May 2015 05:59:33 GMT"
}
] | 2015-05-27T00:00:00 | [
[
"Narasimhan",
"Harikrishna",
""
],
[
"Kar",
"Purushottam",
""
],
[
"Jain",
"Prateek",
""
]
] | TITLE: Optimizing Non-decomposable Performance Measures: A Tale of Two Classes
ABSTRACT: Modern classification problems frequently present mild to severe label
imbalance as well as specific requirements on classification characteristics,
and require optimizing performance measures that are non-decomposable over the
dataset, such as F-measure. Such measures have spurred much interest and pose
specific challenges to learning algorithms since their non-additive nature
precludes a direct application of well-studied large scale optimization methods
such as stochastic gradient descent.
In this paper we reveal that for two large families of performance measures
that can be expressed as functions of true positive/negative rates, it is
indeed possible to implement point stochastic updates. The families we consider
are concave and pseudo-linear functions of TPR, TNR which cover several
popularly used performance measures such as F-measure, G-mean and H-mean.
Our core contribution is an adaptive linearization scheme for these families,
using which we develop optimization techniques that enable truly point-based
stochastic updates. For concave performance measures we propose SPADE, a
stochastic primal dual solver; for pseudo-linear measures we propose STAMP, a
stochastic alternate maximization procedure. Both methods have crisp
convergence guarantees, demonstrate significant speedups over existing methods
- often by an order of magnitude or more, and give similar or more accurate
predictions on test data.
| no_new_dataset | 0.941115 |
1505.06814 | Francesco Palmieri A. N. | Francesco A. N. Palmieri and Amedeo Buonanno | Discrete Independent Component Analysis (DICA) with Belief Propagation | Sumbitted for publication (May 2015) | null | null | null | cs.CV cs.LG stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We apply belief propagation to a Bayesian bipartite graph composed of
discrete independent hidden variables and discrete visible variables. The
network is the Discrete counterpart of Independent Component Analysis (DICA)
and it is manipulated in a factor graph form for inference and learning. A full
set of simulations is reported for character images from the MNIST dataset. The
results show that the factorial code implemented by the sources contributes to
build a good generative model for the data that can be used in various
inference modes.
| [
{
"version": "v1",
"created": "Tue, 26 May 2015 06:02:05 GMT"
}
] | 2015-05-27T00:00:00 | [
[
"Palmieri",
"Francesco A. N.",
""
],
[
"Buonanno",
"Amedeo",
""
]
] | TITLE: Discrete Independent Component Analysis (DICA) with Belief Propagation
ABSTRACT: We apply belief propagation to a Bayesian bipartite graph composed of
discrete independent hidden variables and discrete visible variables. The
network is the Discrete counterpart of Independent Component Analysis (DICA)
and it is manipulated in a factor graph form for inference and learning. A full
set of simulations is reported for character images from the MNIST dataset. The
results show that the factorial code implemented by the sources contributes to
build a good generative model for the data that can be used in various
inference modes.
| no_new_dataset | 0.947381 |
1505.06907 | Andreas Gr\"unauer | Andreas Gr\"unauer and Markus Vincze | Using Dimension Reduction to Improve the Classification of
High-dimensional Data | Presented at OAGM Workshop, 2015 (arXiv:1505.01065) | null | null | OAGM/2015/09 | cs.LG cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this work we show that the classification performance of high-dimensional
structural MRI data with only a small set of training examples is improved by
the usage of dimension reduction methods. We assessed two different dimension
reduction variants: feature selection by ANOVA F-test and feature
transformation by PCA. On the reduced datasets, we applied common learning
algorithms using 5-fold cross-validation. Training, tuning of the
hyperparameters, as well as the performance evaluation of the classifiers was
conducted using two different performance measures: Accuracy, and Receiver
Operating Characteristic curve (AUC). Our hypothesis is supported by
experimental results.
| [
{
"version": "v1",
"created": "Tue, 26 May 2015 11:33:04 GMT"
}
] | 2015-05-27T00:00:00 | [
[
"Grünauer",
"Andreas",
""
],
[
"Vincze",
"Markus",
""
]
] | TITLE: Using Dimension Reduction to Improve the Classification of
High-dimensional Data
ABSTRACT: In this work we show that the classification performance of high-dimensional
structural MRI data with only a small set of training examples is improved by
the usage of dimension reduction methods. We assessed two different dimension
reduction variants: feature selection by ANOVA F-test and feature
transformation by PCA. On the reduced datasets, we applied common learning
algorithms using 5-fold cross-validation. Training, tuning of the
hyperparameters, as well as the performance evaluation of the classifiers was
conducted using two different performance measures: Accuracy, and Receiver
Operating Characteristic curve (AUC). Our hypothesis is supported by
experimental results.
| no_new_dataset | 0.953013 |
1505.06915 | Jean-Philippe Vert | K\'evin Vervier (CBIO), Pierre Mah\'e, Maud Tournoud, Jean-Baptiste
Veyrieras, Jean-Philippe Vert (CBIO) | Large-scale Machine Learning for Metagenomics Sequence Classification | null | null | null | null | q-bio.QM cs.CE cs.LG q-bio.GN stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Metagenomics characterizes the taxonomic diversity of microbial communities
by sequencing DNA directly from an environmental sample. One of the main
challenges in metagenomics data analysis is the binning step, where each
sequenced read is assigned to a taxonomic clade. Due to the large volume of
metagenomics datasets, binning methods need fast and accurate algorithms that
can operate with reasonable computing requirements. While standard
alignment-based methods provide state-of-the-art performance, compositional
approaches that assign a taxonomic class to a DNA read based on the k-mers it
contains have the potential to provide faster solutions. In this work, we
investigate the potential of modern, large-scale machine learning
implementations for taxonomic affectation of next-generation sequencing reads
based on their k-mers profile. We show that machine learning-based
compositional approaches benefit from increasing the number of fragments
sampled from reference genome to tune their parameters, up to a coverage of
about 10, and from increasing the k-mer size to about 12. Tuning these models
involves training a machine learning model on about 10 8 samples in 10 7
dimensions, which is out of reach of standard soft-wares but can be done
efficiently with modern implementations for large-scale machine learning. The
resulting models are competitive in terms of accuracy with well-established
alignment tools for problems involving a small to moderate number of candidate
species, and for reasonable amounts of sequencing errors. We show, however,
that compositional approaches are still limited in their ability to deal with
problems involving a greater number of species, and more sensitive to
sequencing errors. We finally confirm that compositional approach achieve
faster prediction times, with a gain of 3 to 15 times with respect to the
BWA-MEM short read mapper, depending on the number of candidate species and the
level of sequencing noise.
| [
{
"version": "v1",
"created": "Tue, 26 May 2015 12:02:04 GMT"
}
] | 2015-05-27T00:00:00 | [
[
"Vervier",
"Kévin",
"",
"CBIO"
],
[
"Mahé",
"Pierre",
"",
"CBIO"
],
[
"Tournoud",
"Maud",
"",
"CBIO"
],
[
"Veyrieras",
"Jean-Baptiste",
"",
"CBIO"
],
[
"Vert",
"Jean-Philippe",
"",
"CBIO"
]
] | TITLE: Large-scale Machine Learning for Metagenomics Sequence Classification
ABSTRACT: Metagenomics characterizes the taxonomic diversity of microbial communities
by sequencing DNA directly from an environmental sample. One of the main
challenges in metagenomics data analysis is the binning step, where each
sequenced read is assigned to a taxonomic clade. Due to the large volume of
metagenomics datasets, binning methods need fast and accurate algorithms that
can operate with reasonable computing requirements. While standard
alignment-based methods provide state-of-the-art performance, compositional
approaches that assign a taxonomic class to a DNA read based on the k-mers it
contains have the potential to provide faster solutions. In this work, we
investigate the potential of modern, large-scale machine learning
implementations for taxonomic affectation of next-generation sequencing reads
based on their k-mers profile. We show that machine learning-based
compositional approaches benefit from increasing the number of fragments
sampled from reference genome to tune their parameters, up to a coverage of
about 10, and from increasing the k-mer size to about 12. Tuning these models
involves training a machine learning model on about 10 8 samples in 10 7
dimensions, which is out of reach of standard soft-wares but can be done
efficiently with modern implementations for large-scale machine learning. The
resulting models are competitive in terms of accuracy with well-established
alignment tools for problems involving a small to moderate number of candidate
species, and for reasonable amounts of sequencing errors. We show, however,
that compositional approaches are still limited in their ability to deal with
problems involving a greater number of species, and more sensitive to
sequencing errors. We finally confirm that compositional approach achieve
faster prediction times, with a gain of 3 to 15 times with respect to the
BWA-MEM short read mapper, depending on the number of candidate species and the
level of sequencing noise.
| no_new_dataset | 0.948394 |
1505.06918 | Roman Lutz | Roman Lutz | Fantasy Football Prediction | class project, 7 pages (1 sources, 1 appendix) | null | null | null | cs.LG | http://creativecommons.org/licenses/by/3.0/ | The ubiquity of professional sports and specifically the NFL have lead to an
increase in popularity for Fantasy Football. Users have many tools at their
disposal: statistics, predictions, rankings of experts and even recommendations
of peers. There are issues with all of these, though. Especially since many
people pay money to play, the prediction tools should be enhanced as they
provide unbiased and easy-to-use assistance for users. This paper provides and
discusses approaches to predict Fantasy Football scores of Quarterbacks with
relatively limited data. In addition to that, it includes several suggestions
on how the data could be enhanced to achieve better results. The dataset
consists only of game data from the last six NFL seasons. I used two different
methods to predict the Fantasy Football scores of NFL players: Support Vector
Regression (SVR) and Neural Networks. The results of both are promising given
the limited data that was used.
| [
{
"version": "v1",
"created": "Tue, 26 May 2015 12:14:56 GMT"
}
] | 2015-05-27T00:00:00 | [
[
"Lutz",
"Roman",
""
]
] | TITLE: Fantasy Football Prediction
ABSTRACT: The ubiquity of professional sports and specifically the NFL have lead to an
increase in popularity for Fantasy Football. Users have many tools at their
disposal: statistics, predictions, rankings of experts and even recommendations
of peers. There are issues with all of these, though. Especially since many
people pay money to play, the prediction tools should be enhanced as they
provide unbiased and easy-to-use assistance for users. This paper provides and
discusses approaches to predict Fantasy Football scores of Quarterbacks with
relatively limited data. In addition to that, it includes several suggestions
on how the data could be enhanced to achieve better results. The dataset
consists only of game data from the last six NFL seasons. I used two different
methods to predict the Fantasy Football scores of NFL players: Support Vector
Regression (SVR) and Neural Networks. The results of both are promising given
the limited data that was used.
| no_new_dataset | 0.950503 |
1505.06999 | Luis Ortiz | Joshua Belanich and Luis E. Ortiz | Some Open Problems in Optimal AdaBoost and Decision Stumps | 4 pages, rejected from COLT15 Open Problems May 19, 2015 (submitted
April 21, 2015; original 3 pages in COLT-conference format) | null | null | null | cs.LG stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The significance of the study of the theoretical and practical properties of
AdaBoost is unquestionable, given its simplicity, wide practical use, and
effectiveness on real-world datasets. Here we present a few open problems
regarding the behavior of "Optimal AdaBoost," a term coined by Rudin,
Daubechies, and Schapire in 2004 to label the simple version of the standard
AdaBoost algorithm in which the weak learner that AdaBoost uses always outputs
the weak classifier with lowest weighted error among the respective hypothesis
class of weak classifiers implicit in the weak learner. We concentrate on the
standard, "vanilla" version of Optimal AdaBoost for binary classification that
results from using an exponential-loss upper bound on the misclassification
training error. We present two types of open problems. One deals with general
weak hypotheses. The other deals with the particular case of decision stumps,
as often and commonly used in practice. Answers to the open problems can have
immediate significant impact to (1) cementing previously established results on
asymptotic convergence properties of Optimal AdaBoost, for finite datasets,
which in turn can be the start to any convergence-rate analysis; (2)
understanding the weak-hypotheses class of effective decision stumps generated
from data, which we have empirically observed to be significantly smaller than
the typically obtained class, as well as the effect on the weak learner's
running time and previously established improved bounds on the generalization
performance of Optimal AdaBoost classifiers; and (3) shedding some light on the
"self control" that AdaBoost tends to exhibit in practice.
| [
{
"version": "v1",
"created": "Tue, 26 May 2015 15:18:33 GMT"
}
] | 2015-05-27T00:00:00 | [
[
"Belanich",
"Joshua",
""
],
[
"Ortiz",
"Luis E.",
""
]
] | TITLE: Some Open Problems in Optimal AdaBoost and Decision Stumps
ABSTRACT: The significance of the study of the theoretical and practical properties of
AdaBoost is unquestionable, given its simplicity, wide practical use, and
effectiveness on real-world datasets. Here we present a few open problems
regarding the behavior of "Optimal AdaBoost," a term coined by Rudin,
Daubechies, and Schapire in 2004 to label the simple version of the standard
AdaBoost algorithm in which the weak learner that AdaBoost uses always outputs
the weak classifier with lowest weighted error among the respective hypothesis
class of weak classifiers implicit in the weak learner. We concentrate on the
standard, "vanilla" version of Optimal AdaBoost for binary classification that
results from using an exponential-loss upper bound on the misclassification
training error. We present two types of open problems. One deals with general
weak hypotheses. The other deals with the particular case of decision stumps,
as often and commonly used in practice. Answers to the open problems can have
immediate significant impact to (1) cementing previously established results on
asymptotic convergence properties of Optimal AdaBoost, for finite datasets,
which in turn can be the start to any convergence-rate analysis; (2)
understanding the weak-hypotheses class of effective decision stumps generated
from data, which we have empirically observed to be significantly smaller than
the typically obtained class, as well as the effect on the weak learner's
running time and previously established improved bounds on the generalization
performance of Optimal AdaBoost classifiers; and (3) shedding some light on the
"self control" that AdaBoost tends to exhibit in practice.
| no_new_dataset | 0.94366 |
1407.3859 | Jeremy Kepner | Jeremy Kepner, Christian Anderson, William Arcand, David Bestor, Bill
Bergeron, Chansup Byun, Matthew Hubbell, Peter Michaleas, Julie Mullen, David
O'Gwynn, Andrew Prout, Albert Reuther, Antonio Rosa, Charles Yee (MIT) | D4M 2.0 Schema: A General Purpose High Performance Schema for the
Accumulo Database | 6 pages; IEEE HPEC 2013 | null | 10.1109/HPEC.2013.6670318 | null | cs.DB astro-ph.IM cs.DC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Non-traditional, relaxed consistency, triple store databases are the backbone
of many web companies (e.g., Google Big Table, Amazon Dynamo, and Facebook
Cassandra). The Apache Accumulo database is a high performance open source
relaxed consistency database that is widely used for government applications.
Obtaining the full benefits of Accumulo requires using novel schemas. The
Dynamic Distributed Dimensional Data Model (D4M)[http://d4m.mit.edu] provides a
uniform mathematical framework based on associative arrays that encompasses
both traditional (i.e., SQL) and non-traditional databases. For non-traditional
databases D4M naturally leads to a general purpose schema that can be used to
fully index and rapidly query every unique string in a dataset. The D4M 2.0
Schema has been applied with little or no customization to cyber,
bioinformatics, scientific citation, free text, and social media data. The D4M
2.0 Schema is simple, requires minimal parsing, and achieves the highest
published Accumulo ingest rates. The benefits of the D4M 2.0 Schema are
independent of the D4M interface. Any interface to Accumulo can achieve these
benefits by using the D4M 2.0 Schema
| [
{
"version": "v1",
"created": "Tue, 15 Jul 2014 01:54:45 GMT"
}
] | 2015-05-26T00:00:00 | [
[
"Kepner",
"Jeremy",
"",
"MIT"
],
[
"Anderson",
"Christian",
"",
"MIT"
],
[
"Arcand",
"William",
"",
"MIT"
],
[
"Bestor",
"David",
"",
"MIT"
],
[
"Bergeron",
"Bill",
"",
"MIT"
],
[
"Byun",
"Chansup",
"",
"MIT"
],
[
"Hubbell",
"Matthew",
"",
"MIT"
],
[
"Michaleas",
"Peter",
"",
"MIT"
],
[
"Mullen",
"Julie",
"",
"MIT"
],
[
"O'Gwynn",
"David",
"",
"MIT"
],
[
"Prout",
"Andrew",
"",
"MIT"
],
[
"Reuther",
"Albert",
"",
"MIT"
],
[
"Rosa",
"Antonio",
"",
"MIT"
],
[
"Yee",
"Charles",
"",
"MIT"
]
] | TITLE: D4M 2.0 Schema: A General Purpose High Performance Schema for the
Accumulo Database
ABSTRACT: Non-traditional, relaxed consistency, triple store databases are the backbone
of many web companies (e.g., Google Big Table, Amazon Dynamo, and Facebook
Cassandra). The Apache Accumulo database is a high performance open source
relaxed consistency database that is widely used for government applications.
Obtaining the full benefits of Accumulo requires using novel schemas. The
Dynamic Distributed Dimensional Data Model (D4M)[http://d4m.mit.edu] provides a
uniform mathematical framework based on associative arrays that encompasses
both traditional (i.e., SQL) and non-traditional databases. For non-traditional
databases D4M naturally leads to a general purpose schema that can be used to
fully index and rapidly query every unique string in a dataset. The D4M 2.0
Schema has been applied with little or no customization to cyber,
bioinformatics, scientific citation, free text, and social media data. The D4M
2.0 Schema is simple, requires minimal parsing, and achieves the highest
published Accumulo ingest rates. The benefits of the D4M 2.0 Schema are
independent of the D4M interface. Any interface to Accumulo can achieve these
benefits by using the D4M 2.0 Schema
| no_new_dataset | 0.949153 |
1412.1740 | Matthew Kusner | Matt J. Kusner, Nicholas I. Kolkin, Stephen Tyree, Kilian Q.
Weinberger | Image Data Compression for Covariance and Histogram Descriptors | null | null | null | null | stat.ML cs.CV cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Covariance and histogram image descriptors provide an effective way to
capture information about images. Both excel when used in combination with
special purpose distance metrics. For covariance descriptors these metrics
measure the distance along the non-Euclidean Riemannian manifold of symmetric
positive definite matrices. For histogram descriptors the Earth Mover's
distance measures the optimal transport between two histograms. Although more
precise, these distance metrics are very expensive to compute, making them
impractical in many applications, even for data sets of only a few thousand
examples. In this paper we present two methods to compress the size of
covariance and histogram datasets with only marginal increases in test error
for k-nearest neighbor classification. Specifically, we show that we can reduce
data sets to 16% and in some cases as little as 2% of their original size,
while approximately matching the test error of kNN classification on the full
training set. In fact, because the compressed set is learned in a supervised
fashion, it sometimes even outperforms the full data set, while requiring only
a fraction of the space and drastically reducing test-time computation.
| [
{
"version": "v1",
"created": "Thu, 4 Dec 2014 17:22:22 GMT"
},
{
"version": "v2",
"created": "Sat, 23 May 2015 17:07:59 GMT"
}
] | 2015-05-26T00:00:00 | [
[
"Kusner",
"Matt J.",
""
],
[
"Kolkin",
"Nicholas I.",
""
],
[
"Tyree",
"Stephen",
""
],
[
"Weinberger",
"Kilian Q.",
""
]
] | TITLE: Image Data Compression for Covariance and Histogram Descriptors
ABSTRACT: Covariance and histogram image descriptors provide an effective way to
capture information about images. Both excel when used in combination with
special purpose distance metrics. For covariance descriptors these metrics
measure the distance along the non-Euclidean Riemannian manifold of symmetric
positive definite matrices. For histogram descriptors the Earth Mover's
distance measures the optimal transport between two histograms. Although more
precise, these distance metrics are very expensive to compute, making them
impractical in many applications, even for data sets of only a few thousand
examples. In this paper we present two methods to compress the size of
covariance and histogram datasets with only marginal increases in test error
for k-nearest neighbor classification. Specifically, we show that we can reduce
data sets to 16% and in some cases as little as 2% of their original size,
while approximately matching the test error of kNN classification on the full
training set. In fact, because the compressed set is learned in a supervised
fashion, it sometimes even outperforms the full data set, while requiring only
a fraction of the space and drastically reducing test-time computation.
| no_new_dataset | 0.950227 |
1504.06580 | Cicero dos Santos | Cicero Nogueira dos Santos, Bing Xiang, Bowen Zhou | Classifying Relations by Ranking with Convolutional Neural Networks | Accepted as a long paper in the 53rd Annual Meeting of the
Association for Computational Linguistics (ACL 2015) | null | null | null | cs.CL cs.LG cs.NE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Relation classification is an important semantic processing task for which
state-ofthe-art systems still rely on costly handcrafted features. In this work
we tackle the relation classification task using a convolutional neural network
that performs classification by ranking (CR-CNN). We propose a new pairwise
ranking loss function that makes it easy to reduce the impact of artificial
classes. We perform experiments using the the SemEval-2010 Task 8 dataset,
which is designed for the task of classifying the relationship between two
nominals marked in a sentence. Using CRCNN, we outperform the state-of-the-art
for this dataset and achieve a F1 of 84.1 without using any costly handcrafted
features. Additionally, our experimental results show that: (1) our approach is
more effective than CNN followed by a softmax classifier; (2) omitting the
representation of the artificial class Other improves both precision and
recall; and (3) using only word embeddings as input features is enough to
achieve state-of-the-art results if we consider only the text between the two
target nominals.
| [
{
"version": "v1",
"created": "Fri, 24 Apr 2015 17:50:33 GMT"
},
{
"version": "v2",
"created": "Sun, 24 May 2015 13:58:05 GMT"
}
] | 2015-05-26T00:00:00 | [
[
"Santos",
"Cicero Nogueira dos",
""
],
[
"Xiang",
"Bing",
""
],
[
"Zhou",
"Bowen",
""
]
] | TITLE: Classifying Relations by Ranking with Convolutional Neural Networks
ABSTRACT: Relation classification is an important semantic processing task for which
state-ofthe-art systems still rely on costly handcrafted features. In this work
we tackle the relation classification task using a convolutional neural network
that performs classification by ranking (CR-CNN). We propose a new pairwise
ranking loss function that makes it easy to reduce the impact of artificial
classes. We perform experiments using the the SemEval-2010 Task 8 dataset,
which is designed for the task of classifying the relationship between two
nominals marked in a sentence. Using CRCNN, we outperform the state-of-the-art
for this dataset and achieve a F1 of 84.1 without using any costly handcrafted
features. Additionally, our experimental results show that: (1) our approach is
more effective than CNN followed by a softmax classifier; (2) omitting the
representation of the artificial class Other improves both precision and
recall; and (3) using only word embeddings as input features is enough to
achieve state-of-the-art results if we consider only the text between the two
target nominals.
| no_new_dataset | 0.946646 |
1505.06249 | Alexandre Drouin | Alexandre Drouin, S\'ebastien Gigu\`ere, Maxime D\'eraspe,
Fran\c{c}ois Laviolette, Mario Marchand, Jacques Corbeil | Greedy Biomarker Discovery in the Genome with Applications to
Antimicrobial Resistance | Peer-reviewed and accepted for an oral presentation in the Greed is
Great workshop at the International Conference on Machine Learning, Lille,
France, 2015 | null | null | null | q-bio.GN cs.LG stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The Set Covering Machine (SCM) is a greedy learning algorithm that produces
sparse classifiers. We extend the SCM for datasets that contain a huge number
of features. The whole genetic material of living organisms is an example of
such a case, where the number of feature exceeds 10^7. Three human pathogens
were used to evaluate the performance of the SCM at predicting antimicrobial
resistance. Our results show that the SCM compares favorably in terms of
sparsity and accuracy against L1 and L2 regularized Support Vector Machines and
CART decision trees. Moreover, the SCM was the only algorithm that could
consider the full feature space. For all other algorithms, the latter had to be
filtered as a preprocessing step.
| [
{
"version": "v1",
"created": "Fri, 22 May 2015 23:29:40 GMT"
}
] | 2015-05-26T00:00:00 | [
[
"Drouin",
"Alexandre",
""
],
[
"Giguère",
"Sébastien",
""
],
[
"Déraspe",
"Maxime",
""
],
[
"Laviolette",
"François",
""
],
[
"Marchand",
"Mario",
""
],
[
"Corbeil",
"Jacques",
""
]
] | TITLE: Greedy Biomarker Discovery in the Genome with Applications to
Antimicrobial Resistance
ABSTRACT: The Set Covering Machine (SCM) is a greedy learning algorithm that produces
sparse classifiers. We extend the SCM for datasets that contain a huge number
of features. The whole genetic material of living organisms is an example of
such a case, where the number of feature exceeds 10^7. Three human pathogens
were used to evaluate the performance of the SCM at predicting antimicrobial
resistance. Our results show that the SCM compares favorably in terms of
sparsity and accuracy against L1 and L2 regularized Support Vector Machines and
CART decision trees. Moreover, the SCM was the only algorithm that could
consider the full feature space. For all other algorithms, the latter had to be
filtered as a preprocessing step.
| no_new_dataset | 0.952309 |
1505.06250 | George Toderici | Balakrishnan Varadarajan and George Toderici and Sudheendra
Vijayanarasimhan and Apostol Natsev | Efficient Large Scale Video Classification | null | null | null | null | cs.CV cs.MM cs.NE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Video classification has advanced tremendously over the recent years. A large
part of the improvements in video classification had to do with the work done
by the image classification community and the use of deep convolutional
networks (CNNs) which produce competitive results with hand- crafted motion
features. These networks were adapted to use video frames in various ways and
have yielded state of the art classification results. We present two methods
that build on this work, and scale it up to work with millions of videos and
hundreds of thousands of classes while maintaining a low computational cost. In
the context of large scale video processing, training CNNs on video frames is
extremely time consuming, due to the large number of frames involved. We
propose to avoid this problem by training CNNs on either YouTube thumbnails or
Flickr images, and then using these networks' outputs as features for other
higher level classifiers. We discuss the challenges of achieving this and
propose two models for frame-level and video-level classification. The first is
a highly efficient mixture of experts while the latter is based on long short
term memory neural networks. We present results on the Sports-1M video dataset
(1 million videos, 487 classes) and on a new dataset which has 12 million
videos and 150,000 labels.
| [
{
"version": "v1",
"created": "Fri, 22 May 2015 23:45:32 GMT"
}
] | 2015-05-26T00:00:00 | [
[
"Varadarajan",
"Balakrishnan",
""
],
[
"Toderici",
"George",
""
],
[
"Vijayanarasimhan",
"Sudheendra",
""
],
[
"Natsev",
"Apostol",
""
]
] | TITLE: Efficient Large Scale Video Classification
ABSTRACT: Video classification has advanced tremendously over the recent years. A large
part of the improvements in video classification had to do with the work done
by the image classification community and the use of deep convolutional
networks (CNNs) which produce competitive results with hand- crafted motion
features. These networks were adapted to use video frames in various ways and
have yielded state of the art classification results. We present two methods
that build on this work, and scale it up to work with millions of videos and
hundreds of thousands of classes while maintaining a low computational cost. In
the context of large scale video processing, training CNNs on video frames is
extremely time consuming, due to the large number of frames involved. We
propose to avoid this problem by training CNNs on either YouTube thumbnails or
Flickr images, and then using these networks' outputs as features for other
higher level classifiers. We discuss the challenges of achieving this and
propose two models for frame-level and video-level classification. The first is
a highly efficient mixture of experts while the latter is based on long short
term memory neural networks. We present results on the Sports-1M video dataset
(1 million videos, 487 classes) and on a new dataset which has 12 million
videos and 150,000 labels.
| new_dataset | 0.962427 |
1505.06485 | Syama Sundar Rangapuram | Syama Sundar Rangapuram and Matthias Hein | Constrained 1-Spectral Clustering | Long version of paper accepted at AISTATS 2012 | null | null | null | stat.ML cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | An important form of prior information in clustering comes in form of
cannot-link and must-link constraints. We present a generalization of the
popular spectral clustering technique which integrates such constraints.
Motivated by the recently proposed $1$-spectral clustering for the
unconstrained problem, our method is based on a tight relaxation of the
constrained normalized cut into a continuous optimization problem. Opposite to
all other methods which have been suggested for constrained spectral
clustering, we can always guarantee to satisfy all constraints. Moreover, our
soft formulation allows to optimize a trade-off between normalized cut and the
number of violated constraints. An efficient implementation is provided which
scales to large datasets. We outperform consistently all other proposed methods
in the experiments.
| [
{
"version": "v1",
"created": "Sun, 24 May 2015 21:25:44 GMT"
}
] | 2015-05-26T00:00:00 | [
[
"Rangapuram",
"Syama Sundar",
""
],
[
"Hein",
"Matthias",
""
]
] | TITLE: Constrained 1-Spectral Clustering
ABSTRACT: An important form of prior information in clustering comes in form of
cannot-link and must-link constraints. We present a generalization of the
popular spectral clustering technique which integrates such constraints.
Motivated by the recently proposed $1$-spectral clustering for the
unconstrained problem, our method is based on a tight relaxation of the
constrained normalized cut into a continuous optimization problem. Opposite to
all other methods which have been suggested for constrained spectral
clustering, we can always guarantee to satisfy all constraints. Moreover, our
soft formulation allows to optimize a trade-off between normalized cut and the
number of violated constraints. An efficient implementation is provided which
scales to large datasets. We outperform consistently all other proposed methods
in the experiments.
| no_new_dataset | 0.944689 |
1505.06531 | Tsu-Wei Chen | Tsu-Wei Chen, Meena Abdelmaseeh, Daniel Stashuk | Affine and Regional Dynamic Time Warpng | null | null | null | null | cs.CV cs.CE cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Pointwise matches between two time series are of great importance in time
series analysis, and dynamic time warping (DTW) is known to provide generally
reasonable matches. There are situations where time series alignment should be
invariant to scaling and offset in amplitude or where local regions of the
considered time series should be strongly reflected in pointwise matches. Two
different variants of DTW, affine DTW (ADTW) and regional DTW (RDTW), are
proposed to handle scaling and offset in amplitude and provide regional
emphasis respectively. Furthermore, ADTW and RDTW can be combined in two
different ways to generate alignments that incorporate advantages from both
methods, where the affine model can be applied either globally to the entire
time series or locally to each region. The proposed alignment methods
outperform DTW on specific simulated datasets, and one-nearest-neighbor
classifiers using their associated difference measures are competitive with the
difference measures associated with state-of-the-art alignment methods on real
datasets.
| [
{
"version": "v1",
"created": "Mon, 25 May 2015 03:23:31 GMT"
}
] | 2015-05-26T00:00:00 | [
[
"Chen",
"Tsu-Wei",
""
],
[
"Abdelmaseeh",
"Meena",
""
],
[
"Stashuk",
"Daniel",
""
]
] | TITLE: Affine and Regional Dynamic Time Warpng
ABSTRACT: Pointwise matches between two time series are of great importance in time
series analysis, and dynamic time warping (DTW) is known to provide generally
reasonable matches. There are situations where time series alignment should be
invariant to scaling and offset in amplitude or where local regions of the
considered time series should be strongly reflected in pointwise matches. Two
different variants of DTW, affine DTW (ADTW) and regional DTW (RDTW), are
proposed to handle scaling and offset in amplitude and provide regional
emphasis respectively. Furthermore, ADTW and RDTW can be combined in two
different ways to generate alignments that incorporate advantages from both
methods, where the affine model can be applied either globally to the entire
time series or locally to each region. The proposed alignment methods
outperform DTW on specific simulated datasets, and one-nearest-neighbor
classifiers using their associated difference measures are competitive with the
difference measures associated with state-of-the-art alignment methods on real
datasets.
| no_new_dataset | 0.953188 |
1505.06532 | Ali Jahanian | Ali Jahanian, S. V. N. Vishwanathan, Jan P. Allebach | Colors $-$Messengers of Concepts: Visual Design Mining for Learning
Color Semantics | null | null | null | null | cs.HC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper studies the concept of color semantics by modeling a dataset of
magazine cover designs, evaluating the model via crowdsourcing, and
demonstrating several prototypes that facilitate color-related design tasks. We
investigate a probabilistic generative modeling framework that expresses
semantic concepts as a combination of color and word distributions
$-$color-word topics. We adopt an extension to Latent Dirichlet Allocation
(LDA) topic modeling called LDA-dual to infer a set of color-word topics over a
corpus of 2,654 magazine covers spanning 71 distinct titles and 12 genres.
While LDA models text documents as distributions over word topics, we model
magazine covers as distributions over color-word topics. The results of our
crowdsourced experiments confirm that the model is able to successfully
discover the associations between colors and linguistic concepts. Finally, we
demonstrate several simple prototypes that apply the learned model to color
palette recommendation, design example retrieval, image retrieval, image color
selection, and image recoloring.
| [
{
"version": "v1",
"created": "Mon, 25 May 2015 03:34:46 GMT"
}
] | 2015-05-26T00:00:00 | [
[
"Jahanian",
"Ali",
""
],
[
"Vishwanathan",
"S. V. N.",
""
],
[
"Allebach",
"Jan P.",
""
]
] | TITLE: Colors $-$Messengers of Concepts: Visual Design Mining for Learning
Color Semantics
ABSTRACT: This paper studies the concept of color semantics by modeling a dataset of
magazine cover designs, evaluating the model via crowdsourcing, and
demonstrating several prototypes that facilitate color-related design tasks. We
investigate a probabilistic generative modeling framework that expresses
semantic concepts as a combination of color and word distributions
$-$color-word topics. We adopt an extension to Latent Dirichlet Allocation
(LDA) topic modeling called LDA-dual to infer a set of color-word topics over a
corpus of 2,654 magazine covers spanning 71 distinct titles and 12 genres.
While LDA models text documents as distributions over word topics, we model
magazine covers as distributions over color-word topics. The results of our
crowdsourced experiments confirm that the model is able to successfully
discover the associations between colors and linguistic concepts. Finally, we
demonstrate several simple prototypes that apply the learned model to color
palette recommendation, design example retrieval, image retrieval, image color
selection, and image recoloring.
| no_new_dataset | 0.943138 |
1505.06538 | J Massey Cashore | J. Massey Cashore, Xiaoting Zhao, Alexander A. Alemi, Yujia Liu, Peter
I. Frazier | Clustering via Content-Augmented Stochastic Blockmodels | null | null | null | null | stat.ML cs.LG cs.SI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Much of the data being created on the web contains interactions between users
and items. Stochastic blockmodels, and other methods for community detection
and clustering of bipartite graphs, can infer latent user communities and
latent item clusters from this interaction data. These methods, however,
typically ignore the items' contents and the information they provide about
item clusters, despite the tendency of items in the same latent cluster to
share commonalities in content. We introduce content-augmented stochastic
blockmodels (CASB), which use item content together with user-item interaction
data to enhance the user communities and item clusters learned. Comparisons to
several state-of-the-art benchmark methods, on datasets arising from scientists
interacting with scientific articles, show that content-augmented stochastic
blockmodels provide highly accurate clusters with respect to metrics
representative of the underlying community structure.
| [
{
"version": "v1",
"created": "Mon, 25 May 2015 04:19:12 GMT"
}
] | 2015-05-26T00:00:00 | [
[
"Cashore",
"J. Massey",
""
],
[
"Zhao",
"Xiaoting",
""
],
[
"Alemi",
"Alexander A.",
""
],
[
"Liu",
"Yujia",
""
],
[
"Frazier",
"Peter I.",
""
]
] | TITLE: Clustering via Content-Augmented Stochastic Blockmodels
ABSTRACT: Much of the data being created on the web contains interactions between users
and items. Stochastic blockmodels, and other methods for community detection
and clustering of bipartite graphs, can infer latent user communities and
latent item clusters from this interaction data. These methods, however,
typically ignore the items' contents and the information they provide about
item clusters, despite the tendency of items in the same latent cluster to
share commonalities in content. We introduce content-augmented stochastic
blockmodels (CASB), which use item content together with user-item interaction
data to enhance the user communities and item clusters learned. Comparisons to
several state-of-the-art benchmark methods, on datasets arising from scientists
interacting with scientific articles, show that content-augmented stochastic
blockmodels provide highly accurate clusters with respect to metrics
representative of the underlying community structure.
| no_new_dataset | 0.946794 |
1505.05921 | Katherine Driggs-Campbell | Katherine Driggs-Campbell and Ruzena Bajcsy | Identifying Modes of Intent from Driver Behaviors in Dynamic
Environments | Submitted to ITSC 2015 | null | null | null | cs.SY | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In light of growing attention of intelligent vehicle systems, we propose
developing a driver model that uses a hybrid system formulation to capture the
intent of the driver. This model hopes to capture human driving behavior in a
way that can be utilized by semi- and fully autonomous systems in heterogeneous
environments. We consider a discrete set of high level goals or intent modes,
that is designed to encompass the decision making process of the human. A
driver model is derived using a dataset of lane changes collected in a
realistic driving simulator, in which the driver actively labels data to give
us insight into her intent. By building the labeled dataset, we are able to
utilize classification tools to build the driver model using features of based
on her perception of the environment, and achieve high accuracy in identifying
driver intent. Multiple algorithms are presented and compared on the dataset,
and a comparison of the varying behaviors between drivers is drawn. Using this
modeling methodology, we present a model that can be used to assess driver
behaviors and to develop human-inspired safety metrics that can be utilized in
intelligent vehicular systems.
| [
{
"version": "v1",
"created": "Thu, 21 May 2015 23:19:09 GMT"
}
] | 2015-05-25T00:00:00 | [
[
"Driggs-Campbell",
"Katherine",
""
],
[
"Bajcsy",
"Ruzena",
""
]
] | TITLE: Identifying Modes of Intent from Driver Behaviors in Dynamic
Environments
ABSTRACT: In light of growing attention of intelligent vehicle systems, we propose
developing a driver model that uses a hybrid system formulation to capture the
intent of the driver. This model hopes to capture human driving behavior in a
way that can be utilized by semi- and fully autonomous systems in heterogeneous
environments. We consider a discrete set of high level goals or intent modes,
that is designed to encompass the decision making process of the human. A
driver model is derived using a dataset of lane changes collected in a
realistic driving simulator, in which the driver actively labels data to give
us insight into her intent. By building the labeled dataset, we are able to
utilize classification tools to build the driver model using features of based
on her perception of the environment, and achieve high accuracy in identifying
driver intent. Multiple algorithms are presented and compared on the dataset,
and a comparison of the varying behaviors between drivers is drawn. Using this
modeling methodology, we present a model that can be used to assess driver
behaviors and to develop human-inspired safety metrics that can be utilized in
intelligent vehicular systems.
| new_dataset | 0.9598 |
1505.05957 | Tianmin Shu | Tianmin Shu, Dan Xie, Brandon Rothrock, Sinisa Todorovic, Song-Chun
Zhu | Joint Inference of Groups, Events and Human Roles in Aerial Videos | CVPR 2015 Oral Presentation | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | With the advent of drones, aerial video analysis becomes increasingly
important; yet, it has received scant attention in the literature. This paper
addresses a new problem of parsing low-resolution aerial videos of large
spatial areas, in terms of 1) grouping, 2) recognizing events and 3) assigning
roles to people engaged in events. We propose a novel framework aimed at
conducting joint inference of the above tasks, as reasoning about each in
isolation typically fails in our setting. Given noisy tracklets of people and
detections of large objects and scene surfaces (e.g., building, grass), we use
a spatiotemporal AND-OR graph to drive our joint inference, using Markov Chain
Monte Carlo and dynamic programming. We also introduce a new formalism of
spatiotemporal templates characterizing latent sub-events. For evaluation, we
have collected and released a new aerial videos dataset using a hex-rotor
flying over picnic areas rich with group events. Our results demonstrate that
we successfully address above inference tasks under challenging conditions.
| [
{
"version": "v1",
"created": "Fri, 22 May 2015 05:59:18 GMT"
}
] | 2015-05-25T00:00:00 | [
[
"Shu",
"Tianmin",
""
],
[
"Xie",
"Dan",
""
],
[
"Rothrock",
"Brandon",
""
],
[
"Todorovic",
"Sinisa",
""
],
[
"Zhu",
"Song-Chun",
""
]
] | TITLE: Joint Inference of Groups, Events and Human Roles in Aerial Videos
ABSTRACT: With the advent of drones, aerial video analysis becomes increasingly
important; yet, it has received scant attention in the literature. This paper
addresses a new problem of parsing low-resolution aerial videos of large
spatial areas, in terms of 1) grouping, 2) recognizing events and 3) assigning
roles to people engaged in events. We propose a novel framework aimed at
conducting joint inference of the above tasks, as reasoning about each in
isolation typically fails in our setting. Given noisy tracklets of people and
detections of large objects and scene surfaces (e.g., building, grass), we use
a spatiotemporal AND-OR graph to drive our joint inference, using Markov Chain
Monte Carlo and dynamic programming. We also introduce a new formalism of
spatiotemporal templates characterizing latent sub-events. For evaluation, we
have collected and released a new aerial videos dataset using a hex-rotor
flying over picnic areas rich with group events. Our results demonstrate that
we successfully address above inference tasks under challenging conditions.
| new_dataset | 0.954942 |
1505.06169 | Emma Strubell | Emma Strubell, Luke Vilnis, Kate Silverstein, Andrew McCallum | Learning Dynamic Feature Selection for Fast Sequential Prediction | Appears in The 53rd Annual Meeting of the Association for
Computational Linguistics, Beijing, China, July 2015 | null | null | null | cs.CL cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We present paired learning and inference algorithms for significantly
reducing computation and increasing speed of the vector dot products in the
classifiers that are at the heart of many NLP components. This is accomplished
by partitioning the features into a sequence of templates which are ordered
such that high confidence can often be reached using only a small fraction of
all features. Parameter estimation is arranged to maximize accuracy and early
confidence in this sequence. Our approach is simpler and better suited to NLP
than other related cascade methods. We present experiments in left-to-right
part-of-speech tagging, named entity recognition, and transition-based
dependency parsing. On the typical benchmarking datasets we can preserve POS
tagging accuracy above 97% and parsing LAS above 88.5% both with over a
five-fold reduction in run-time, and NER F1 above 88 with more than 2x increase
in speed.
| [
{
"version": "v1",
"created": "Fri, 22 May 2015 18:28:21 GMT"
}
] | 2015-05-25T00:00:00 | [
[
"Strubell",
"Emma",
""
],
[
"Vilnis",
"Luke",
""
],
[
"Silverstein",
"Kate",
""
],
[
"McCallum",
"Andrew",
""
]
] | TITLE: Learning Dynamic Feature Selection for Fast Sequential Prediction
ABSTRACT: We present paired learning and inference algorithms for significantly
reducing computation and increasing speed of the vector dot products in the
classifiers that are at the heart of many NLP components. This is accomplished
by partitioning the features into a sequence of templates which are ordered
such that high confidence can often be reached using only a small fraction of
all features. Parameter estimation is arranged to maximize accuracy and early
confidence in this sequence. Our approach is simpler and better suited to NLP
than other related cascade methods. We present experiments in left-to-right
part-of-speech tagging, named entity recognition, and transition-based
dependency parsing. On the typical benchmarking datasets we can preserve POS
tagging accuracy above 97% and parsing LAS above 88.5% both with over a
five-fold reduction in run-time, and NER F1 above 88 with more than 2x increase
in speed.
| no_new_dataset | 0.945751 |
1401.4489 | Devansh Arpit | Devansh Arpit, Ifeoma Nwogu, Gaurav Srivastava, Venu Govindaraju | An Analysis of Random Projections in Cancelable Biometrics | null | null | null | null | cs.CV cs.LG stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | With increasing concerns about security, the need for highly secure physical
biometrics-based authentication systems utilizing \emph{cancelable biometric}
technologies is on the rise. Because the problem of cancelable template
generation deals with the trade-off between template security and matching
performance, many state-of-the-art algorithms successful in generating high
quality cancelable biometrics all have random projection as one of their early
processing steps. This paper therefore presents a formal analysis of why random
projections is an essential step in cancelable biometrics. By formally defining
the notion of an \textit{Independent Subspace Structure} for datasets, it can
be shown that random projection preserves the subspace structure of data
vectors generated from a union of independent linear subspaces. The bound on
the minimum number of random vectors required for this to hold is also derived
and is shown to depend logarithmically on the number of data samples, not only
in independent subspaces but in disjoint subspace settings as well. The
theoretical analysis presented is supported in detail with empirical results on
real-world face recognition datasets.
| [
{
"version": "v1",
"created": "Fri, 17 Jan 2014 23:21:56 GMT"
},
{
"version": "v2",
"created": "Wed, 5 Feb 2014 02:57:25 GMT"
},
{
"version": "v3",
"created": "Fri, 14 Nov 2014 02:38:09 GMT"
}
] | 2015-05-22T00:00:00 | [
[
"Arpit",
"Devansh",
""
],
[
"Nwogu",
"Ifeoma",
""
],
[
"Srivastava",
"Gaurav",
""
],
[
"Govindaraju",
"Venu",
""
]
] | TITLE: An Analysis of Random Projections in Cancelable Biometrics
ABSTRACT: With increasing concerns about security, the need for highly secure physical
biometrics-based authentication systems utilizing \emph{cancelable biometric}
technologies is on the rise. Because the problem of cancelable template
generation deals with the trade-off between template security and matching
performance, many state-of-the-art algorithms successful in generating high
quality cancelable biometrics all have random projection as one of their early
processing steps. This paper therefore presents a formal analysis of why random
projections is an essential step in cancelable biometrics. By formally defining
the notion of an \textit{Independent Subspace Structure} for datasets, it can
be shown that random projection preserves the subspace structure of data
vectors generated from a union of independent linear subspaces. The bound on
the minimum number of random vectors required for this to hold is also derived
and is shown to depend logarithmically on the number of data samples, not only
in independent subspaces but in disjoint subspace settings as well. The
theoretical analysis presented is supported in detail with empirical results on
real-world face recognition datasets.
| no_new_dataset | 0.948251 |
1407.7073 | Weinan Zhang | Weinan Zhang, Shuai Yuan, Jun Wang, Xuehua Shen | Real-Time Bidding Benchmarking with iPinYou Dataset | UCL Technical Report 2014 | null | null | null | cs.GT cs.CY | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Being an emerging paradigm for display advertising, Real-Time Bidding (RTB)
drives the focus of the bidding strategy from context to users' interest by
computing a bid for each impression in real time. The data mining work and
particularly the bidding strategy development becomes crucial in this
performance-driven business. However, researchers in computational advertising
area have been suffering from lack of publicly available benchmark datasets,
which are essential to compare different algorithms and systems. Fortunately, a
leading Chinese advertising technology company iPinYou decided to release the
dataset used in its global RTB algorithm competition in 2013. The dataset
includes logs of ad auctions, bids, impressions, clicks, and final conversions.
These logs reflect the market environment as well as form a complete path of
users' responses from advertisers' perspective. This dataset directly supports
the experiments of some important research problems such as bid optimisation
and CTR estimation. To the best of our knowledge, this is the first publicly
available dataset on RTB display advertising. Thus, they are valuable for
reproducible research and understanding the whole RTB ecosystem. In this paper,
we first provide the detailed statistical analysis of this dataset. Then we
introduce the research problem of bid optimisation in RTB and the simple yet
comprehensive evaluation protocol. Besides, a series of benchmark experiments
are also conducted, including both click-through rate (CTR) estimation and bid
optimisation.
| [
{
"version": "v1",
"created": "Fri, 25 Jul 2014 23:20:29 GMT"
},
{
"version": "v2",
"created": "Fri, 1 Aug 2014 11:22:17 GMT"
},
{
"version": "v3",
"created": "Thu, 21 May 2015 18:20:30 GMT"
}
] | 2015-05-22T00:00:00 | [
[
"Zhang",
"Weinan",
""
],
[
"Yuan",
"Shuai",
""
],
[
"Wang",
"Jun",
""
],
[
"Shen",
"Xuehua",
""
]
] | TITLE: Real-Time Bidding Benchmarking with iPinYou Dataset
ABSTRACT: Being an emerging paradigm for display advertising, Real-Time Bidding (RTB)
drives the focus of the bidding strategy from context to users' interest by
computing a bid for each impression in real time. The data mining work and
particularly the bidding strategy development becomes crucial in this
performance-driven business. However, researchers in computational advertising
area have been suffering from lack of publicly available benchmark datasets,
which are essential to compare different algorithms and systems. Fortunately, a
leading Chinese advertising technology company iPinYou decided to release the
dataset used in its global RTB algorithm competition in 2013. The dataset
includes logs of ad auctions, bids, impressions, clicks, and final conversions.
These logs reflect the market environment as well as form a complete path of
users' responses from advertisers' perspective. This dataset directly supports
the experiments of some important research problems such as bid optimisation
and CTR estimation. To the best of our knowledge, this is the first publicly
available dataset on RTB display advertising. Thus, they are valuable for
reproducible research and understanding the whole RTB ecosystem. In this paper,
we first provide the detailed statistical analysis of this dataset. Then we
introduce the research problem of bid optimisation in RTB and the simple yet
comprehensive evaluation protocol. Besides, a series of benchmark experiments
are also conducted, including both click-through rate (CTR) estimation and bid
optimisation.
| new_dataset | 0.96793 |
1502.01602 | Donn Morrison | V\'aclav Bel\'ak, Afra Mashhadi, Alessandra Sala, Donn Morrison | Phantom cascades: The effect of hidden nodes on information diffusion | Preprint submitted to Elsevier Computer Communications | null | null | null | cs.SI physics.soc-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Research on information diffusion generally assumes complete knowledge of the
underlying network. However, in the presence of factors such as increasing
privacy awareness, restrictions on application programming interfaces (APIs)
and sampling strategies, this assumption rarely holds in the real world which
in turn leads to an underestimation of the size of information cascades. In
this work we study the effect of hidden network structure on information
diffusion processes. We characterise information cascades through activation
paths traversing visible and hidden parts of the network. We quantify diffusion
estimation error while varying the amount of hidden structure in five empirical
and synthetic network datasets and demonstrate the effect of topological
properties on this error. Finally, we suggest practical recommendations for
practitioners and propose a model to predict the cascade size with minimal
information regarding the underlying network.
| [
{
"version": "v1",
"created": "Thu, 5 Feb 2015 15:13:33 GMT"
},
{
"version": "v2",
"created": "Thu, 21 May 2015 13:34:14 GMT"
}
] | 2015-05-22T00:00:00 | [
[
"Belák",
"Václav",
""
],
[
"Mashhadi",
"Afra",
""
],
[
"Sala",
"Alessandra",
""
],
[
"Morrison",
"Donn",
""
]
] | TITLE: Phantom cascades: The effect of hidden nodes on information diffusion
ABSTRACT: Research on information diffusion generally assumes complete knowledge of the
underlying network. However, in the presence of factors such as increasing
privacy awareness, restrictions on application programming interfaces (APIs)
and sampling strategies, this assumption rarely holds in the real world which
in turn leads to an underestimation of the size of information cascades. In
this work we study the effect of hidden network structure on information
diffusion processes. We characterise information cascades through activation
paths traversing visible and hidden parts of the network. We quantify diffusion
estimation error while varying the amount of hidden structure in five empirical
and synthetic network datasets and demonstrate the effect of topological
properties on this error. Finally, we suggest practical recommendations for
practitioners and propose a model to predict the cascade size with minimal
information regarding the underlying network.
| no_new_dataset | 0.951233 |
1503.00756 | Marco Stronati | Konstantinos Chatzikokolakis, Catuscia Palamidessi, Marco Stronati | Constructing elastic distinguishability metrics for location privacy | null | null | 10.1515/popets-2015-0023 | null | cs.CR | http://creativecommons.org/licenses/by/3.0/ | With the increasing popularity of hand-held devices, location-based
applications and services have access to accurate and real-time location
information, raising serious privacy concerns for their users. The recently
introduced notion of geo-indistinguishability tries to address this problem by
adapting the well-known concept of differential privacy to the area of
location-based systems. Although geo-indistinguishability presents various
appealing aspects, it has the problem of treating space in a uniform way,
imposing the addition of the same amount of noise everywhere on the map. In
this paper we propose a novel elastic distinguishability metric that warps the
geometrical distance, capturing the different degrees of density of each area.
As a consequence, the obtained mechanism adapts the level of noise while
achieving the same degree of privacy everywhere. We also show how such an
elastic metric can easily incorporate the concept of a "geographic fence" that
is commonly employed to protect the highly recurrent locations of a user, such
as his home or work. We perform an extensive evaluation of our technique by
building an elastic metric for Paris' wide metropolitan area, using semantic
information from the OpenStreetMap database. We compare the resulting mechanism
against the Planar Laplace mechanism satisfying standard
geo-indistinguishability, using two real-world datasets from the Gowalla and
Brightkite location-based social networks. The results show that the elastic
mechanism adapts well to the semantics of each area, adjusting the noise as we
move outside the city center, hence offering better overall privacy.
| [
{
"version": "v1",
"created": "Mon, 2 Mar 2015 21:32:11 GMT"
},
{
"version": "v2",
"created": "Thu, 21 May 2015 09:39:47 GMT"
}
] | 2015-05-22T00:00:00 | [
[
"Chatzikokolakis",
"Konstantinos",
""
],
[
"Palamidessi",
"Catuscia",
""
],
[
"Stronati",
"Marco",
""
]
] | TITLE: Constructing elastic distinguishability metrics for location privacy
ABSTRACT: With the increasing popularity of hand-held devices, location-based
applications and services have access to accurate and real-time location
information, raising serious privacy concerns for their users. The recently
introduced notion of geo-indistinguishability tries to address this problem by
adapting the well-known concept of differential privacy to the area of
location-based systems. Although geo-indistinguishability presents various
appealing aspects, it has the problem of treating space in a uniform way,
imposing the addition of the same amount of noise everywhere on the map. In
this paper we propose a novel elastic distinguishability metric that warps the
geometrical distance, capturing the different degrees of density of each area.
As a consequence, the obtained mechanism adapts the level of noise while
achieving the same degree of privacy everywhere. We also show how such an
elastic metric can easily incorporate the concept of a "geographic fence" that
is commonly employed to protect the highly recurrent locations of a user, such
as his home or work. We perform an extensive evaluation of our technique by
building an elastic metric for Paris' wide metropolitan area, using semantic
information from the OpenStreetMap database. We compare the resulting mechanism
against the Planar Laplace mechanism satisfying standard
geo-indistinguishability, using two real-world datasets from the Gowalla and
Brightkite location-based social networks. The results show that the elastic
mechanism adapts well to the semantics of each area, adjusting the noise as we
move outside the city center, hence offering better overall privacy.
| no_new_dataset | 0.949248 |
1505.05667 | Xipeng Qiu | Chenxi Zhu, Xipeng Qiu, Xinchi Chen, Xuanjing Huang | A Re-ranking Model for Dependency Parser with Recursive Convolutional
Neural Network | null | null | null | null | cs.CL cs.LG cs.NE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this work, we address the problem to model all the nodes (words or
phrases) in a dependency tree with the dense representations. We propose a
recursive convolutional neural network (RCNN) architecture to capture syntactic
and compositional-semantic representations of phrases and words in a dependency
tree. Different with the original recursive neural network, we introduce the
convolution and pooling layers, which can model a variety of compositions by
the feature maps and choose the most informative compositions by the pooling
layers. Based on RCNN, we use a discriminative model to re-rank a $k$-best list
of candidate dependency parsing trees. The experiments show that RCNN is very
effective to improve the state-of-the-art dependency parsing on both English
and Chinese datasets.
| [
{
"version": "v1",
"created": "Thu, 21 May 2015 10:23:10 GMT"
}
] | 2015-05-22T00:00:00 | [
[
"Zhu",
"Chenxi",
""
],
[
"Qiu",
"Xipeng",
""
],
[
"Chen",
"Xinchi",
""
],
[
"Huang",
"Xuanjing",
""
]
] | TITLE: A Re-ranking Model for Dependency Parser with Recursive Convolutional
Neural Network
ABSTRACT: In this work, we address the problem to model all the nodes (words or
phrases) in a dependency tree with the dense representations. We propose a
recursive convolutional neural network (RCNN) architecture to capture syntactic
and compositional-semantic representations of phrases and words in a dependency
tree. Different with the original recursive neural network, we introduce the
convolution and pooling layers, which can model a variety of compositions by
the feature maps and choose the most informative compositions by the pooling
layers. Based on RCNN, we use a discriminative model to re-rank a $k$-best list
of candidate dependency parsing trees. The experiments show that RCNN is very
effective to improve the state-of-the-art dependency parsing on both English
and Chinese datasets.
| no_new_dataset | 0.950595 |
1505.05723 | Indre Zliobaite | Indre Zliobaite | On the relation between accuracy and fairness in binary classification | Accepted for presentation to the 2nd workshop on Fairness,
Accountability, and Transparency in Machine Learning (http://www.fatml.org/) | null | null | null | cs.LG cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Our study revisits the problem of accuracy-fairness tradeoff in binary
classification. We argue that comparison of non-discriminatory classifiers
needs to account for different rates of positive predictions, otherwise
conclusions about performance may be misleading, because accuracy and
discrimination of naive baselines on the same dataset vary with different rates
of positive predictions. We provide methodological recommendations for sound
comparison of non-discriminatory classifiers, and present a brief theoretical
and empirical analysis of tradeoffs between accuracy and non-discrimination.
| [
{
"version": "v1",
"created": "Thu, 21 May 2015 13:20:06 GMT"
}
] | 2015-05-22T00:00:00 | [
[
"Zliobaite",
"Indre",
""
]
] | TITLE: On the relation between accuracy and fairness in binary classification
ABSTRACT: Our study revisits the problem of accuracy-fairness tradeoff in binary
classification. We argue that comparison of non-discriminatory classifiers
needs to account for different rates of positive predictions, otherwise
conclusions about performance may be misleading, because accuracy and
discrimination of naive baselines on the same dataset vary with different rates
of positive predictions. We provide methodological recommendations for sound
comparison of non-discriminatory classifiers, and present a brief theoretical
and empirical analysis of tradeoffs between accuracy and non-discrimination.
| no_new_dataset | 0.944331 |
1505.05753 | Mario Fritz | Iaroslav Shcherbatyi, Andreas Bulling, Mario Fritz | GazeDPM: Early Integration of Gaze Information in Deformable Part Models | null | null | null | null | cs.CV cs.HC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | An increasing number of works explore collaborative human-computer systems in
which human gaze is used to enhance computer vision systems. For object
detection these efforts were so far restricted to late integration approaches
that have inherent limitations, such as increased precision without increase in
recall. We propose an early integration approach in a deformable part model,
which constitutes a joint formulation over gaze and visual data. We show that
our GazeDPM method improves over the state-of-the-art DPM baseline by 4% and a
recent method for gaze-supported object detection by 3% on the public POET
dataset. Our approach additionally provides introspection of the learnt models,
can reveal salient image structures, and allows us to investigate the interplay
between gaze attracting and repelling areas, the importance of view-specific
models, as well as viewers' personal biases in gaze patterns. We finally study
important practical aspects of our approach, such as the impact of using
saliency maps instead of real fixations, the impact of the number of fixations,
as well as robustness to gaze estimation error.
| [
{
"version": "v1",
"created": "Thu, 21 May 2015 14:39:51 GMT"
}
] | 2015-05-22T00:00:00 | [
[
"Shcherbatyi",
"Iaroslav",
""
],
[
"Bulling",
"Andreas",
""
],
[
"Fritz",
"Mario",
""
]
] | TITLE: GazeDPM: Early Integration of Gaze Information in Deformable Part Models
ABSTRACT: An increasing number of works explore collaborative human-computer systems in
which human gaze is used to enhance computer vision systems. For object
detection these efforts were so far restricted to late integration approaches
that have inherent limitations, such as increased precision without increase in
recall. We propose an early integration approach in a deformable part model,
which constitutes a joint formulation over gaze and visual data. We show that
our GazeDPM method improves over the state-of-the-art DPM baseline by 4% and a
recent method for gaze-supported object detection by 3% on the public POET
dataset. Our approach additionally provides introspection of the learnt models,
can reveal salient image structures, and allows us to investigate the interplay
between gaze attracting and repelling areas, the importance of view-specific
models, as well as viewers' personal biases in gaze patterns. We finally study
important practical aspects of our approach, such as the impact of using
saliency maps instead of real fixations, the impact of the number of fixations,
as well as robustness to gaze estimation error.
| no_new_dataset | 0.946695 |
1409.4450 | Giovanni Luca Ciampaglia | Giovanni Luca Ciampaglia, Alessandro Flammini, Filippo Menczer | The production of information in the attention economy | 14 pages, 3 figures, 1 table | null | 10.1038/srep09452 | Giovanni Luca Ciampaglia, Alessandro Flammini & Filippo Menczer
Scientific Reports 5, Article number: 9452 (2015) | physics.soc-ph cs.SI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Online traces of human activity offer novel opportunities to study the
dynamics of complex knowledge exchange networks, and in particular how the
relationship between demand and supply of information is mediated by
competition for our limited individual attention. The emergent patterns of
collective attention determine what new information is generated and consumed.
Can we measure the relationship between demand and supply for new information
about a topic? Here we propose a normalization method to compare attention
bursts statistics across topics that have an heterogeneous distribution of
attention. Through analysis of a massive dataset on traffic to Wikipedia, we
find that the production of new knowledge is associated to significant shifts
of collective attention, which we take as a proxy for its demand. What we
observe is consistent with a scenario in which the allocation of attention
toward a topic stimulates the demand for information about it, and in turn the
supply of further novel information. Our attempt to quantify demand and supply
of information, and our finding about their temporal ordering, may lead to the
development of the fundamental laws of the attention economy, and a better
understanding of the social exchange of knowledge in online and offline
information networks.
| [
{
"version": "v1",
"created": "Mon, 15 Sep 2014 21:13:35 GMT"
}
] | 2015-05-21T00:00:00 | [
[
"Ciampaglia",
"Giovanni Luca",
""
],
[
"Flammini",
"Alessandro",
""
],
[
"Menczer",
"Filippo",
""
]
] | TITLE: The production of information in the attention economy
ABSTRACT: Online traces of human activity offer novel opportunities to study the
dynamics of complex knowledge exchange networks, and in particular how the
relationship between demand and supply of information is mediated by
competition for our limited individual attention. The emergent patterns of
collective attention determine what new information is generated and consumed.
Can we measure the relationship between demand and supply for new information
about a topic? Here we propose a normalization method to compare attention
bursts statistics across topics that have an heterogeneous distribution of
attention. Through analysis of a massive dataset on traffic to Wikipedia, we
find that the production of new knowledge is associated to significant shifts
of collective attention, which we take as a proxy for its demand. What we
observe is consistent with a scenario in which the allocation of attention
toward a topic stimulates the demand for information about it, and in turn the
supply of further novel information. Our attempt to quantify demand and supply
of information, and our finding about their temporal ordering, may lead to the
development of the fundamental laws of the attention economy, and a better
understanding of the social exchange of knowledge in online and offline
information networks.
| no_new_dataset | 0.936981 |
1502.04623 | Ivo Danihelka | Karol Gregor, Ivo Danihelka, Alex Graves, Danilo Jimenez Rezende, Daan
Wierstra | DRAW: A Recurrent Neural Network For Image Generation | null | null | null | null | cs.CV cs.LG cs.NE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper introduces the Deep Recurrent Attentive Writer (DRAW) neural
network architecture for image generation. DRAW networks combine a novel
spatial attention mechanism that mimics the foveation of the human eye, with a
sequential variational auto-encoding framework that allows for the iterative
construction of complex images. The system substantially improves on the state
of the art for generative models on MNIST, and, when trained on the Street View
House Numbers dataset, it generates images that cannot be distinguished from
real data with the naked eye.
| [
{
"version": "v1",
"created": "Mon, 16 Feb 2015 16:48:56 GMT"
},
{
"version": "v2",
"created": "Wed, 20 May 2015 15:29:42 GMT"
}
] | 2015-05-21T00:00:00 | [
[
"Gregor",
"Karol",
""
],
[
"Danihelka",
"Ivo",
""
],
[
"Graves",
"Alex",
""
],
[
"Rezende",
"Danilo Jimenez",
""
],
[
"Wierstra",
"Daan",
""
]
] | TITLE: DRAW: A Recurrent Neural Network For Image Generation
ABSTRACT: This paper introduces the Deep Recurrent Attentive Writer (DRAW) neural
network architecture for image generation. DRAW networks combine a novel
spatial attention mechanism that mimics the foveation of the human eye, with a
sequential variational auto-encoding framework that allows for the iterative
construction of complex images. The system substantially improves on the state
of the art for generative models on MNIST, and, when trained on the Street View
House Numbers dataset, it generates images that cannot be distinguished from
real data with the naked eye.
| no_new_dataset | 0.953057 |
1504.06755 | Pingmei Xu | Pingmei Xu, Krista A Ehinger, Yinda Zhang, Adam Finkelstein, Sanjeev
R. Kulkarni, Jianxiong Xiao | TurkerGaze: Crowdsourcing Saliency with Webcam based Eye Tracking | 9 pages, 14 figures | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Traditional eye tracking requires specialized hardware, which means
collecting gaze data from many observers is expensive, tedious and slow.
Therefore, existing saliency prediction datasets are order-of-magnitudes
smaller than typical datasets for other vision recognition tasks. The small
size of these datasets limits the potential for training data intensive
algorithms, and causes overfitting in benchmark evaluation. To address this
deficiency, this paper introduces a webcam-based gaze tracking system that
supports large-scale, crowdsourced eye tracking deployed on Amazon Mechanical
Turk (AMTurk). By a combination of careful algorithm and gaming protocol
design, our system obtains eye tracking data for saliency prediction comparable
to data gathered in a traditional lab setting, with relatively lower cost and
less effort on the part of the researchers. Using this tool, we build a
saliency dataset for a large number of natural images. We will open-source our
tool and provide a web server where researchers can upload their images to get
eye tracking results from AMTurk.
| [
{
"version": "v1",
"created": "Sat, 25 Apr 2015 19:26:47 GMT"
},
{
"version": "v2",
"created": "Wed, 20 May 2015 18:51:23 GMT"
}
] | 2015-05-21T00:00:00 | [
[
"Xu",
"Pingmei",
""
],
[
"Ehinger",
"Krista A",
""
],
[
"Zhang",
"Yinda",
""
],
[
"Finkelstein",
"Adam",
""
],
[
"Kulkarni",
"Sanjeev R.",
""
],
[
"Xiao",
"Jianxiong",
""
]
] | TITLE: TurkerGaze: Crowdsourcing Saliency with Webcam based Eye Tracking
ABSTRACT: Traditional eye tracking requires specialized hardware, which means
collecting gaze data from many observers is expensive, tedious and slow.
Therefore, existing saliency prediction datasets are order-of-magnitudes
smaller than typical datasets for other vision recognition tasks. The small
size of these datasets limits the potential for training data intensive
algorithms, and causes overfitting in benchmark evaluation. To address this
deficiency, this paper introduces a webcam-based gaze tracking system that
supports large-scale, crowdsourced eye tracking deployed on Amazon Mechanical
Turk (AMTurk). By a combination of careful algorithm and gaming protocol
design, our system obtains eye tracking data for saliency prediction comparable
to data gathered in a traditional lab setting, with relatively lower cost and
less effort on the part of the researchers. Using this tool, we build a
saliency dataset for a large number of natural images. We will open-source our
tool and provide a web server where researchers can upload their images to get
eye tracking results from AMTurk.
| new_dataset | 0.947137 |
1505.05208 | Raajen Patel | Raajen Patel, Thomas A. Goldstein, Eva L. Dyer, Azalia Mirhoseini, and
Richard G. Baraniuk | oASIS: Adaptive Column Sampling for Kernel Matrix Approximation | null | null | null | null | stat.ML cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Kernel matrices (e.g. Gram or similarity matrices) are essential for many
state-of-the-art approaches to classification, clustering, and dimensionality
reduction. For large datasets, the cost of forming and factoring such kernel
matrices becomes intractable. To address this challenge, we introduce a new
adaptive sampling algorithm called Accelerated Sequential Incoherence Selection
(oASIS) that samples columns without explicitly computing the entire kernel
matrix. We provide conditions under which oASIS is guaranteed to exactly
recover the kernel matrix with an optimal number of columns selected. Numerical
experiments on both synthetic and real-world datasets demonstrate that oASIS
achieves performance comparable to state-of-the-art adaptive sampling methods
at a fraction of the computational cost. The low runtime complexity of oASIS
and its low memory footprint enable the solution of large problems that are
simply intractable using other adaptive methods.
| [
{
"version": "v1",
"created": "Tue, 19 May 2015 23:12:01 GMT"
}
] | 2015-05-21T00:00:00 | [
[
"Patel",
"Raajen",
""
],
[
"Goldstein",
"Thomas A.",
""
],
[
"Dyer",
"Eva L.",
""
],
[
"Mirhoseini",
"Azalia",
""
],
[
"Baraniuk",
"Richard G.",
""
]
] | TITLE: oASIS: Adaptive Column Sampling for Kernel Matrix Approximation
ABSTRACT: Kernel matrices (e.g. Gram or similarity matrices) are essential for many
state-of-the-art approaches to classification, clustering, and dimensionality
reduction. For large datasets, the cost of forming and factoring such kernel
matrices becomes intractable. To address this challenge, we introduce a new
adaptive sampling algorithm called Accelerated Sequential Incoherence Selection
(oASIS) that samples columns without explicitly computing the entire kernel
matrix. We provide conditions under which oASIS is guaranteed to exactly
recover the kernel matrix with an optimal number of columns selected. Numerical
experiments on both synthetic and real-world datasets demonstrate that oASIS
achieves performance comparable to state-of-the-art adaptive sampling methods
at a fraction of the computational cost. The low runtime complexity of oASIS
and its low memory footprint enable the solution of large problems that are
simply intractable using other adaptive methods.
| no_new_dataset | 0.944485 |
1505.05211 | Souvik Bhattacherjee | Souvik Bhattacherjee and Amit Chavan and Silu Huang and Amol Deshpande
and Aditya Parameswaran | Principles of Dataset Versioning: Exploring the Recreation/Storage
Tradeoff | null | null | null | null | cs.DB | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The relative ease of collaborative data science and analysis has led to a
proliferation of many thousands or millions of $versions$ of the same datasets
in many scientific and commercial domains, acquired or constructed at various
stages of data analysis across many users, and often over long periods of time.
Managing, storing, and recreating these dataset versions is a non-trivial task.
The fundamental challenge here is the $storage-recreation\;trade-off$: the more
storage we use, the faster it is to recreate or retrieve versions, while the
less storage we use, the slower it is to recreate or retrieve versions. Despite
the fundamental nature of this problem, there has been a surprisingly little
amount of work on it. In this paper, we study this trade-off in a principled
manner: we formulate six problems under various settings, trading off these
quantities in various ways, demonstrate that most of the problems are
intractable, and propose a suite of inexpensive heuristics drawing from
techniques in delay-constrained scheduling, and spanning tree literature, to
solve these problems. We have built a prototype version management system, that
aims to serve as a foundation to our DATAHUB system for facilitating
collaborative data science. We demonstrate, via extensive experiments, that our
proposed heuristics provide efficient solutions in practical dataset versioning
scenarios.
| [
{
"version": "v1",
"created": "Tue, 19 May 2015 23:45:05 GMT"
}
] | 2015-05-21T00:00:00 | [
[
"Bhattacherjee",
"Souvik",
""
],
[
"Chavan",
"Amit",
""
],
[
"Huang",
"Silu",
""
],
[
"Deshpande",
"Amol",
""
],
[
"Parameswaran",
"Aditya",
""
]
] | TITLE: Principles of Dataset Versioning: Exploring the Recreation/Storage
Tradeoff
ABSTRACT: The relative ease of collaborative data science and analysis has led to a
proliferation of many thousands or millions of $versions$ of the same datasets
in many scientific and commercial domains, acquired or constructed at various
stages of data analysis across many users, and often over long periods of time.
Managing, storing, and recreating these dataset versions is a non-trivial task.
The fundamental challenge here is the $storage-recreation\;trade-off$: the more
storage we use, the faster it is to recreate or retrieve versions, while the
less storage we use, the slower it is to recreate or retrieve versions. Despite
the fundamental nature of this problem, there has been a surprisingly little
amount of work on it. In this paper, we study this trade-off in a principled
manner: we formulate six problems under various settings, trading off these
quantities in various ways, demonstrate that most of the problems are
intractable, and propose a suite of inexpensive heuristics drawing from
techniques in delay-constrained scheduling, and spanning tree literature, to
solve these problems. We have built a prototype version management system, that
aims to serve as a foundation to our DATAHUB system for facilitating
collaborative data science. We demonstrate, via extensive experiments, that our
proposed heuristics provide efficient solutions in practical dataset versioning
scenarios.
| no_new_dataset | 0.944382 |
1505.05225 | Lihua Guo | Guo Lihua, Li Fudi | Image aesthetic evaluation using paralleled deep convolution neural
network | 7 pages, 6 figures, 9 tables | null | null | null | cs.CV cs.MM | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Image aesthetic evaluation has attracted much attention in recent years.
Image aesthetic evaluation methods heavily depend on the effective aesthetic
feature. Traditional meth-ods always extract hand-crafted features. However,
these hand-crafted features are always designed to adapt particu-lar datasets,
and extraction of them needs special design. Rather than extracting
hand-crafted features, an automati-cally learn of aesthetic features based on
deep convolutional neural network (DCNN) is first adopt in this paper. As we
all know, when the training dataset is given, the DCNN architecture with high
complexity may meet the over-fitting problem. On the other side, the DCNN
architecture with low complexity would not efficiently extract effective
features. For these reasons, we further propose a paralleled convolutional
neural network (PDCNN) with multi-level structures to automatically adapt to
the training dataset. Experimental results show that our proposed PDCNN
architecture achieves better performance than other traditional methods.
| [
{
"version": "v1",
"created": "Wed, 20 May 2015 02:03:23 GMT"
}
] | 2015-05-21T00:00:00 | [
[
"Lihua",
"Guo",
""
],
[
"Fudi",
"Li",
""
]
] | TITLE: Image aesthetic evaluation using paralleled deep convolution neural
network
ABSTRACT: Image aesthetic evaluation has attracted much attention in recent years.
Image aesthetic evaluation methods heavily depend on the effective aesthetic
feature. Traditional meth-ods always extract hand-crafted features. However,
these hand-crafted features are always designed to adapt particu-lar datasets,
and extraction of them needs special design. Rather than extracting
hand-crafted features, an automati-cally learn of aesthetic features based on
deep convolutional neural network (DCNN) is first adopt in this paper. As we
all know, when the training dataset is given, the DCNN architecture with high
complexity may meet the over-fitting problem. On the other side, the DCNN
architecture with low complexity would not efficiently extract effective
features. For these reasons, we further propose a paralleled convolutional
neural network (PDCNN) with multi-level structures to automatically adapt to
the training dataset. Experimental results show that our proposed PDCNN
architecture achieves better performance than other traditional methods.
| no_new_dataset | 0.951549 |
1505.05232 | Songfan Yang | Songfan Yang, Deva Ramanan | Multi-scale recognition with DAG-CNNs | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We explore multi-scale convolutional neural nets (CNNs) for image
classification. Contemporary approaches extract features from a single output
layer. By extracting features from multiple layers, one can simultaneously
reason about high, mid, and low-level features during classification. The
resulting multi-scale architecture can itself be seen as a feed-forward model
that is structured as a directed acyclic graph (DAG-CNNs). We use DAG-CNNs to
learn a set of multiscale features that can be effectively shared between
coarse and fine-grained classification tasks. While fine-tuning such models
helps performance, we show that even "off-the-self" multiscale features perform
quite well. We present extensive analysis and demonstrate state-of-the-art
classification performance on three standard scene benchmarks (SUN397, MIT67,
and Scene15). In terms of the heavily benchmarked MIT67 and Scene15 datasets,
our results reduce the lowest previously-reported error by 23.9% and 9.5%,
respectively.
| [
{
"version": "v1",
"created": "Wed, 20 May 2015 02:52:07 GMT"
}
] | 2015-05-21T00:00:00 | [
[
"Yang",
"Songfan",
""
],
[
"Ramanan",
"Deva",
""
]
] | TITLE: Multi-scale recognition with DAG-CNNs
ABSTRACT: We explore multi-scale convolutional neural nets (CNNs) for image
classification. Contemporary approaches extract features from a single output
layer. By extracting features from multiple layers, one can simultaneously
reason about high, mid, and low-level features during classification. The
resulting multi-scale architecture can itself be seen as a feed-forward model
that is structured as a directed acyclic graph (DAG-CNNs). We use DAG-CNNs to
learn a set of multiscale features that can be effectively shared between
coarse and fine-grained classification tasks. While fine-tuning such models
helps performance, we show that even "off-the-self" multiscale features perform
quite well. We present extensive analysis and demonstrate state-of-the-art
classification performance on three standard scene benchmarks (SUN397, MIT67,
and Scene15). In terms of the heavily benchmarked MIT67 and Scene15 datasets,
our results reduce the lowest previously-reported error by 23.9% and 9.5%,
respectively.
| no_new_dataset | 0.947817 |
1505.05322 | Ai Munandar Tb | Azhari SN and Tb. Ai Munandar | Unsupervised Neural Network-Naive Bayes Model for Grouping Data Regional
Development Results | 6 pages | International Journal of Computer Applications, Volume 104, No 15,
October 2014 | null | null | cs.CY | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Determination quadrant development has an important role in order to
determine the achievement of the development of a district, in terms of the
sector's gross regional domestic product (GDP). The process of determining the
quadrant development typically uses Klassen rules based on its sector GDP. This
study aims to provide a new approach in the conduct of regional development
quadrant clustering using cluster techniques. Clustering is performed based on
the average value of the growth and development of a district contribution
compared with the average value and contribution of the development of the
province based on data in comparison with a year of data to be compared.
Testing models of clustering, performed on a dataset of two provinces, namely
Banten (as a data testing) and Central Java (as the training data), to see the
accuracy of the classification model proposed. The proposed model consists of
two learning methods in it, namely unsupervised (Self Organizing Map / SOM-NN)
method and supervised (Naive Bayess). SOM-NN method is used as a learning
engine to generate training data for the target Class that will be used in the
machine learning Naive Bayess. The results showed the clustering accuracy rate
of the model was 98.1%, while the clustering accuracy rate of the model results
compared to manual analysis shows the accuracy of the typology Klassen smaller,
ie 29.63%. On one side, clustering results of the proposed model is influenced
by the number and keagaraman data sets used.
| [
{
"version": "v1",
"created": "Wed, 20 May 2015 11:22:18 GMT"
}
] | 2015-05-21T00:00:00 | [
[
"SN",
"Azhari",
""
],
[
"Munandar",
"Tb. Ai",
""
]
] | TITLE: Unsupervised Neural Network-Naive Bayes Model for Grouping Data Regional
Development Results
ABSTRACT: Determination quadrant development has an important role in order to
determine the achievement of the development of a district, in terms of the
sector's gross regional domestic product (GDP). The process of determining the
quadrant development typically uses Klassen rules based on its sector GDP. This
study aims to provide a new approach in the conduct of regional development
quadrant clustering using cluster techniques. Clustering is performed based on
the average value of the growth and development of a district contribution
compared with the average value and contribution of the development of the
province based on data in comparison with a year of data to be compared.
Testing models of clustering, performed on a dataset of two provinces, namely
Banten (as a data testing) and Central Java (as the training data), to see the
accuracy of the classification model proposed. The proposed model consists of
two learning methods in it, namely unsupervised (Self Organizing Map / SOM-NN)
method and supervised (Naive Bayess). SOM-NN method is used as a learning
engine to generate training data for the target Class that will be used in the
machine learning Naive Bayess. The results showed the clustering accuracy rate
of the model was 98.1%, while the clustering accuracy rate of the model results
compared to manual analysis shows the accuracy of the typology Klassen smaller,
ie 29.63%. On one side, clustering results of the proposed model is influenced
by the number and keagaraman data sets used.
| no_new_dataset | 0.954351 |
1505.05338 | Poorna Dasgupta | Poorna Banerjee Dasgupta | Algorithmic Analysis of Edge Ranking and Profiling for MTF Determination
of an Imaging System | 3 pages, Published with International Journal of Computer Trends and
Technology (IJCTT), Volume-23 Number-1, 2015 | International Journal of Computer Trends and Technology (IJCTT)
V23(1):46-48, May 2015 | 10.14445/22312803/IJCTT-V23P110 | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Edge detection is one of the most principal techniques for detecting
discontinuities in the gray levels of image pixels. The Modulation Transfer
Function (MTF) is one of the main criteria for assessing imaging quality and is
a parameter frequently used for measuring the sharpness of an imaging system.
In order to determine the MTF, it is essential to determine the best edge from
the target image so that an edge profile can be developed and then the line
spread function and hence the MTF, can be computed accordingly. For regular
image sizes, the human visual system is adept enough to identify suitable edges
from the image. But considering huge image datasets, such as those obtained
from satellites, the image size may range in few gigabytes and in such a case,
manual inspection of images for determination of the best suitable edge is not
plausible and hence, edge profiling tasks have to be automated. This paper
presents a novel, yet simple, algorithm for edge ranking and detection from
image data-sets for MTF computation, which is ideal for automation on
vectorised graphical processing units.
| [
{
"version": "v1",
"created": "Wed, 20 May 2015 12:12:48 GMT"
}
] | 2015-05-21T00:00:00 | [
[
"Dasgupta",
"Poorna Banerjee",
""
]
] | TITLE: Algorithmic Analysis of Edge Ranking and Profiling for MTF Determination
of an Imaging System
ABSTRACT: Edge detection is one of the most principal techniques for detecting
discontinuities in the gray levels of image pixels. The Modulation Transfer
Function (MTF) is one of the main criteria for assessing imaging quality and is
a parameter frequently used for measuring the sharpness of an imaging system.
In order to determine the MTF, it is essential to determine the best edge from
the target image so that an edge profile can be developed and then the line
spread function and hence the MTF, can be computed accordingly. For regular
image sizes, the human visual system is adept enough to identify suitable edges
from the image. But considering huge image datasets, such as those obtained
from satellites, the image size may range in few gigabytes and in such a case,
manual inspection of images for determination of the best suitable edge is not
plausible and hence, edge profiling tasks have to be automated. This paper
presents a novel, yet simple, algorithm for edge ranking and detection from
image data-sets for MTF computation, which is ideal for automation on
vectorised graphical processing units.
| no_new_dataset | 0.947817 |
1505.05354 | Lianwen Jin | Weixin Yang, Lianwen Jin, Dacheng Tao, Zecheng Xie, Ziyong Feng | DropSample: A New Training Method to Enhance Deep Convolutional Neural
Networks for Large-Scale Unconstrained Handwritten Chinese Character
Recognition | 18 pages, 8 figures, 5 tables | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Inspired by the theory of Leitners learning box from the field of psychology,
we propose DropSample, a new method for training deep convolutional neural
networks (DCNNs), and apply it to large-scale online handwritten Chinese
character recognition (HCCR). According to the principle of DropSample, each
training sample is associated with a quota function that is dynamically
adjusted on the basis of the classification confidence given by the DCNN
softmax output. After a learning iteration, samples with low confidence will
have a higher probability of being selected as training data in the next
iteration; in contrast, well-trained and well-recognized samples with very high
confidence will have a lower probability of being involved in the next training
iteration and can be gradually eliminated. As a result, the learning process
becomes more efficient as it progresses. Furthermore, we investigate the use of
domain-specific knowledge to enhance the performance of DCNN by adding a domain
knowledge layer before the traditional CNN. By adopting DropSample together
with different types of domain-specific knowledge, the accuracy of HCCR can be
improved efficiently. Experiments on the CASIA-OLHDWB 1.0, CASIA-OLHWDB 1.1,
and ICDAR 2013 online HCCR competition datasets yield outstanding recognition
rates of 97.33%, 97.06%, and 97.51% respectively, all of which are
significantly better than the previous best results reported in the literature.
| [
{
"version": "v1",
"created": "Wed, 20 May 2015 13:08:57 GMT"
}
] | 2015-05-21T00:00:00 | [
[
"Yang",
"Weixin",
""
],
[
"Jin",
"Lianwen",
""
],
[
"Tao",
"Dacheng",
""
],
[
"Xie",
"Zecheng",
""
],
[
"Feng",
"Ziyong",
""
]
] | TITLE: DropSample: A New Training Method to Enhance Deep Convolutional Neural
Networks for Large-Scale Unconstrained Handwritten Chinese Character
Recognition
ABSTRACT: Inspired by the theory of Leitners learning box from the field of psychology,
we propose DropSample, a new method for training deep convolutional neural
networks (DCNNs), and apply it to large-scale online handwritten Chinese
character recognition (HCCR). According to the principle of DropSample, each
training sample is associated with a quota function that is dynamically
adjusted on the basis of the classification confidence given by the DCNN
softmax output. After a learning iteration, samples with low confidence will
have a higher probability of being selected as training data in the next
iteration; in contrast, well-trained and well-recognized samples with very high
confidence will have a lower probability of being involved in the next training
iteration and can be gradually eliminated. As a result, the learning process
becomes more efficient as it progresses. Furthermore, we investigate the use of
domain-specific knowledge to enhance the performance of DCNN by adding a domain
knowledge layer before the traditional CNN. By adopting DropSample together
with different types of domain-specific knowledge, the accuracy of HCCR can be
improved efficiently. Experiments on the CASIA-OLHDWB 1.0, CASIA-OLHWDB 1.1,
and ICDAR 2013 online HCCR competition datasets yield outstanding recognition
rates of 97.33%, 97.06%, and 97.51% respectively, all of which are
significantly better than the previous best results reported in the literature.
| no_new_dataset | 0.948585 |
1010.1015 | Keith Wiley | Keith Wiley, Andrew Connolly, Jeff Gardner, Simon Krughof, Magdalena
Balazinska, Bill Howe, YongChul Kwon and YingYi Bu | Astronomy in the Cloud: Using MapReduce for Image Coaddition | 31 pages, 11 figures, 2 tables | null | 10.1086/658877 | null | cs.DC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In the coming decade, astronomical surveys of the sky will generate tens of
terabytes of images and detect hundreds of millions of sources every night. The
study of these sources will involve computation challenges such as anomaly
detection and classification, and moving object tracking. Since such studies
benefit from the highest quality data, methods such as image coaddition
(stacking) will be a critical preprocessing step prior to scientific
investigation. With a requirement that these images be analyzed on a nightly
basis to identify moving sources or transient objects, these data streams
present many computational challenges. Given the quantity of data involved, the
computational load of these problems can only be addressed by distributing the
workload over a large number of nodes. However, the high data throughput
demanded by these applications may present scalability challenges for certain
storage architectures. One scalable data-processing method that has emerged in
recent years is MapReduce, and in this paper we focus on its popular
open-source implementation called Hadoop. In the Hadoop framework, the data is
partitioned among storage attached directly to worker nodes, and the processing
workload is scheduled in parallel on the nodes that contain the required input
data. A further motivation for using Hadoop is that it allows us to exploit
cloud computing resources, e.g., Amazon's EC2. We report on our experience
implementing a scalable image-processing pipeline for the SDSS imaging database
using Hadoop. This multi-terabyte imaging dataset provides a good testbed for
algorithm development since its scope and structure approximate future surveys.
First, we describe MapReduce and how we adapted image coaddition to the
MapReduce framework. Then we describe a number of optimizations to our basic
approach and report experimental results comparing their performance.
| [
{
"version": "v1",
"created": "Tue, 5 Oct 2010 20:35:53 GMT"
}
] | 2015-05-20T00:00:00 | [
[
"Wiley",
"Keith",
""
],
[
"Connolly",
"Andrew",
""
],
[
"Gardner",
"Jeff",
""
],
[
"Krughof",
"Simon",
""
],
[
"Balazinska",
"Magdalena",
""
],
[
"Howe",
"Bill",
""
],
[
"Kwon",
"YongChul",
""
],
[
"Bu",
"YingYi",
""
]
] | TITLE: Astronomy in the Cloud: Using MapReduce for Image Coaddition
ABSTRACT: In the coming decade, astronomical surveys of the sky will generate tens of
terabytes of images and detect hundreds of millions of sources every night. The
study of these sources will involve computation challenges such as anomaly
detection and classification, and moving object tracking. Since such studies
benefit from the highest quality data, methods such as image coaddition
(stacking) will be a critical preprocessing step prior to scientific
investigation. With a requirement that these images be analyzed on a nightly
basis to identify moving sources or transient objects, these data streams
present many computational challenges. Given the quantity of data involved, the
computational load of these problems can only be addressed by distributing the
workload over a large number of nodes. However, the high data throughput
demanded by these applications may present scalability challenges for certain
storage architectures. One scalable data-processing method that has emerged in
recent years is MapReduce, and in this paper we focus on its popular
open-source implementation called Hadoop. In the Hadoop framework, the data is
partitioned among storage attached directly to worker nodes, and the processing
workload is scheduled in parallel on the nodes that contain the required input
data. A further motivation for using Hadoop is that it allows us to exploit
cloud computing resources, e.g., Amazon's EC2. We report on our experience
implementing a scalable image-processing pipeline for the SDSS imaging database
using Hadoop. This multi-terabyte imaging dataset provides a good testbed for
algorithm development since its scope and structure approximate future surveys.
First, we describe MapReduce and how we adapted image coaddition to the
MapReduce framework. Then we describe a number of optimizations to our basic
approach and report experimental results comparing their performance.
| no_new_dataset | 0.938857 |
1011.6268 | Marija Mitrovic | Marija Mitrovi\'c, Georgios Paltoglou and Bosiljka Tadi\'c | Quantitative Analysis of Bloggers Collective Behavior Powered by
Emotions | null | null | 10.1088/1742-5468/2011/02/P02005 | 0143821 | physics.soc-ph cond-mat.stat-mech cs.SI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Large-scale data resulting from users online interactions provide the
ultimate source of information to study emergent social phenomena on the Web.
From individual actions of users to observable collective behaviors, different
mechanisms involving emotions expressed in the posted text play a role. Here we
combine approaches of statistical physics with machine-learning methods of text
analysis to study emergence of the emotional behavior among Web users. Mapping
the high-resolution data from digg.com onto bipartite network of users and
their comments onto posted stories, we identify user communities centered
around certain popular posts and determine emotional contents of the related
comments by the emotion-classifier developed for this type of texts. Applied
over different time periods, this framework reveals strong correlations between
the excess of negative emotions and the evolution of communities. We observe
avalanches of emotional comments exhibiting significant self-organized critical
behavior and temporal correlations. To explore robustness of these critical
states, we design a network automaton model on realistic network connections
and several control parameters, which can be inferred from the dataset.
Dissemination of emotions by a small fraction of very active users appears to
critically tune the collective states.
| [
{
"version": "v1",
"created": "Mon, 29 Nov 2010 15:36:39 GMT"
}
] | 2015-05-20T00:00:00 | [
[
"Mitrović",
"Marija",
""
],
[
"Paltoglou",
"Georgios",
""
],
[
"Tadić",
"Bosiljka",
""
]
] | TITLE: Quantitative Analysis of Bloggers Collective Behavior Powered by
Emotions
ABSTRACT: Large-scale data resulting from users online interactions provide the
ultimate source of information to study emergent social phenomena on the Web.
From individual actions of users to observable collective behaviors, different
mechanisms involving emotions expressed in the posted text play a role. Here we
combine approaches of statistical physics with machine-learning methods of text
analysis to study emergence of the emotional behavior among Web users. Mapping
the high-resolution data from digg.com onto bipartite network of users and
their comments onto posted stories, we identify user communities centered
around certain popular posts and determine emotional contents of the related
comments by the emotion-classifier developed for this type of texts. Applied
over different time periods, this framework reveals strong correlations between
the excess of negative emotions and the evolution of communities. We observe
avalanches of emotional comments exhibiting significant self-organized critical
behavior and temporal correlations. To explore robustness of these critical
states, we design a network automaton model on realistic network connections
and several control parameters, which can be inferred from the dataset.
Dissemination of emotions by a small fraction of very active users appears to
critically tune the collective states.
| no_new_dataset | 0.940079 |
1012.1184 | Lei Zhang Dr. | Weisheng Dong, Lei Zhang, Guangming Shi, Xiaolin Wu | Image Deblurring and Super-resolution by Adaptive Sparse Domain
Selection and Adaptive Regularization | 35 pages. This paper is under review in IEEE TIP | null | 10.1109/TIP.2011.2108306 | null | cs.CV cs.MM | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | As a powerful statistical image modeling technique, sparse representation has
been successfully used in various image restoration applications. The success
of sparse representation owes to the development of l1-norm optimization
techniques, and the fact that natural images are intrinsically sparse in some
domain. The image restoration quality largely depends on whether the employed
sparse domain can represent well the underlying image. Considering that the
contents can vary significantly across different images or different patches in
a single image, we propose to learn various sets of bases from a pre-collected
dataset of example image patches, and then for a given patch to be processed,
one set of bases are adaptively selected to characterize the local sparse
domain. We further introduce two adaptive regularization terms into the sparse
representation framework. First, a set of autoregressive (AR) models are
learned from the dataset of example image patches. The best fitted AR models to
a given patch are adaptively selected to regularize the image local structures.
Second, the image non-local self-similarity is introduced as another
regularization term. In addition, the sparsity regularization parameter is
adaptively estimated for better image restoration performance. Extensive
experiments on image deblurring and super-resolution validate that by using
adaptive sparse domain selection and adaptive regularization, the proposed
method achieves much better results than many state-of-the-art algorithms in
terms of both PSNR and visual perception.
| [
{
"version": "v1",
"created": "Mon, 6 Dec 2010 14:37:14 GMT"
}
] | 2015-05-20T00:00:00 | [
[
"Dong",
"Weisheng",
""
],
[
"Zhang",
"Lei",
""
],
[
"Shi",
"Guangming",
""
],
[
"Wu",
"Xiaolin",
""
]
] | TITLE: Image Deblurring and Super-resolution by Adaptive Sparse Domain
Selection and Adaptive Regularization
ABSTRACT: As a powerful statistical image modeling technique, sparse representation has
been successfully used in various image restoration applications. The success
of sparse representation owes to the development of l1-norm optimization
techniques, and the fact that natural images are intrinsically sparse in some
domain. The image restoration quality largely depends on whether the employed
sparse domain can represent well the underlying image. Considering that the
contents can vary significantly across different images or different patches in
a single image, we propose to learn various sets of bases from a pre-collected
dataset of example image patches, and then for a given patch to be processed,
one set of bases are adaptively selected to characterize the local sparse
domain. We further introduce two adaptive regularization terms into the sparse
representation framework. First, a set of autoregressive (AR) models are
learned from the dataset of example image patches. The best fitted AR models to
a given patch are adaptively selected to regularize the image local structures.
Second, the image non-local self-similarity is introduced as another
regularization term. In addition, the sparsity regularization parameter is
adaptively estimated for better image restoration performance. Extensive
experiments on image deblurring and super-resolution validate that by using
adaptive sparse domain selection and adaptive regularization, the proposed
method achieves much better results than many state-of-the-art algorithms in
terms of both PSNR and visual perception.
| no_new_dataset | 0.948632 |
1401.4529 | Bal\'azs Hidasi | Bal\'azs Hidasi, Domonkos Tikk | General factorization framework for context-aware recommendations | The final publication is available at Springer via
http://dx.doi.org/10.1007/s10618-015-0417-y. Data Mining and Knowledge
Discovery, 2015 | null | 10.1007/s10618-015-0417-y | null | cs.IR cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Context-aware recommendation algorithms focus on refining recommendations by
considering additional information, available to the system. This topic has
gained a lot of attention recently. Among others, several factorization methods
were proposed to solve the problem, although most of them assume explicit
feedback which strongly limits their real-world applicability. While these
algorithms apply various loss functions and optimization strategies, the
preference modeling under context is less explored due to the lack of tools
allowing for easy experimentation with various models. As context dimensions
are introduced beyond users and items, the space of possible preference models
and the importance of proper modeling largely increases.
In this paper we propose a General Factorization Framework (GFF), a single
flexible algorithm that takes the preference model as an input and computes
latent feature matrices for the input dimensions. GFF allows us to easily
experiment with various linear models on any context-aware recommendation task,
be it explicit or implicit feedback based. The scaling properties makes it
usable under real life circumstances as well.
We demonstrate the framework's potential by exploring various preference
models on a 4-dimensional context-aware problem with contexts that are
available for almost any real life datasets. We show in our experiments --
performed on five real life, implicit feedback datasets -- that proper
preference modelling significantly increases recommendation accuracy, and
previously unused models outperform the traditional ones. Novel models in GFF
also outperform state-of-the-art factorization algorithms.
We also extend the method to be fully compliant to the Multidimensional
Dataspace Model, one of the most extensive data models of context-enriched
data. Extended GFF allows the seamless incorporation of information into the
fac[truncated]
| [
{
"version": "v1",
"created": "Sat, 18 Jan 2014 11:13:26 GMT"
},
{
"version": "v2",
"created": "Tue, 19 May 2015 11:50:22 GMT"
}
] | 2015-05-20T00:00:00 | [
[
"Hidasi",
"Balázs",
""
],
[
"Tikk",
"Domonkos",
""
]
] | TITLE: General factorization framework for context-aware recommendations
ABSTRACT: Context-aware recommendation algorithms focus on refining recommendations by
considering additional information, available to the system. This topic has
gained a lot of attention recently. Among others, several factorization methods
were proposed to solve the problem, although most of them assume explicit
feedback which strongly limits their real-world applicability. While these
algorithms apply various loss functions and optimization strategies, the
preference modeling under context is less explored due to the lack of tools
allowing for easy experimentation with various models. As context dimensions
are introduced beyond users and items, the space of possible preference models
and the importance of proper modeling largely increases.
In this paper we propose a General Factorization Framework (GFF), a single
flexible algorithm that takes the preference model as an input and computes
latent feature matrices for the input dimensions. GFF allows us to easily
experiment with various linear models on any context-aware recommendation task,
be it explicit or implicit feedback based. The scaling properties makes it
usable under real life circumstances as well.
We demonstrate the framework's potential by exploring various preference
models on a 4-dimensional context-aware problem with contexts that are
available for almost any real life datasets. We show in our experiments --
performed on five real life, implicit feedback datasets -- that proper
preference modelling significantly increases recommendation accuracy, and
previously unused models outperform the traditional ones. Novel models in GFF
also outperform state-of-the-art factorization algorithms.
We also extend the method to be fully compliant to the Multidimensional
Dataspace Model, one of the most extensive data models of context-enriched
data. Extended GFF allows the seamless incorporation of information into the
fac[truncated]
| no_new_dataset | 0.947672 |
1408.6865 | Paul Dauncey | P. D. Dauncey, M. Kenzie, N. Wardle and G. J. Davies | Handling uncertainties in background shapes: the discrete profiling
method | Accepted by J.Inst | null | 10.1088/1748-0221/10/04/P04015 | null | physics.data-an hep-ex | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | A common problem in data analysis is that the functional form, as well as the
parameter values, of the underlying model which should describe a dataset is
not known a priori. In these cases some extra uncertainty must be assigned to
the extracted parameters of interest due to lack of exact knowledge of the
functional form of the model. A method for assigning an appropriate error is
presented. The method is based on considering the choice of functional form as
a discrete nuisance parameter which is profiled in an analogous way to
continuous nuisance parameters. The bias and coverage of this method are shown
to be good when applied to a realistic example.
| [
{
"version": "v1",
"created": "Thu, 28 Aug 2014 21:31:13 GMT"
},
{
"version": "v2",
"created": "Mon, 22 Dec 2014 17:05:11 GMT"
},
{
"version": "v3",
"created": "Wed, 24 Dec 2014 15:13:18 GMT"
},
{
"version": "v4",
"created": "Fri, 20 Feb 2015 10:55:12 GMT"
},
{
"version": "v5",
"created": "Tue, 28 Apr 2015 12:58:45 GMT"
}
] | 2015-05-20T00:00:00 | [
[
"Dauncey",
"P. D.",
""
],
[
"Kenzie",
"M.",
""
],
[
"Wardle",
"N.",
""
],
[
"Davies",
"G. J.",
""
]
] | TITLE: Handling uncertainties in background shapes: the discrete profiling
method
ABSTRACT: A common problem in data analysis is that the functional form, as well as the
parameter values, of the underlying model which should describe a dataset is
not known a priori. In these cases some extra uncertainty must be assigned to
the extracted parameters of interest due to lack of exact knowledge of the
functional form of the model. A method for assigning an appropriate error is
presented. The method is based on considering the choice of functional form as
a discrete nuisance parameter which is profiled in an analogous way to
continuous nuisance parameters. The bias and coverage of this method are shown
to be good when applied to a realistic example.
| no_new_dataset | 0.947186 |
1501.07320 | Arun Tejasvi Chaganty | Volodymyr Kuleshov and Arun Tejasvi Chaganty and Percy Liang | Tensor Factorization via Matrix Factorization | Appearing in Proceedings of the 18th International Conference on
Artificial Intelligence and Statistics (AISTATS) 2015, San Diego, CA, USA.
JMLR: W&CP volume 38 | null | null | null | cs.LG stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Tensor factorization arises in many machine learning applications, such
knowledge base modeling and parameter estimation in latent variable models.
However, numerical methods for tensor factorization have not reached the level
of maturity of matrix factorization methods. In this paper, we propose a new
method for CP tensor factorization that uses random projections to reduce the
problem to simultaneous matrix diagonalization. Our method is conceptually
simple and also applies to non-orthogonal and asymmetric tensors of arbitrary
order. We prove that a small number random projections essentially preserves
the spectral information in the tensor, allowing us to remove the dependence on
the eigengap that plagued earlier tensor-to-matrix reductions. Experimentally,
our method outperforms existing tensor factorization methods on both simulated
data and two real datasets.
| [
{
"version": "v1",
"created": "Thu, 29 Jan 2015 01:01:34 GMT"
},
{
"version": "v2",
"created": "Mon, 18 May 2015 21:53:34 GMT"
}
] | 2015-05-20T00:00:00 | [
[
"Kuleshov",
"Volodymyr",
""
],
[
"Chaganty",
"Arun Tejasvi",
""
],
[
"Liang",
"Percy",
""
]
] | TITLE: Tensor Factorization via Matrix Factorization
ABSTRACT: Tensor factorization arises in many machine learning applications, such
knowledge base modeling and parameter estimation in latent variable models.
However, numerical methods for tensor factorization have not reached the level
of maturity of matrix factorization methods. In this paper, we propose a new
method for CP tensor factorization that uses random projections to reduce the
problem to simultaneous matrix diagonalization. Our method is conceptually
simple and also applies to non-orthogonal and asymmetric tensors of arbitrary
order. We prove that a small number random projections essentially preserves
the spectral information in the tensor, allowing us to remove the dependence on
the eigengap that plagued earlier tensor-to-matrix reductions. Experimentally,
our method outperforms existing tensor factorization methods on both simulated
data and two real datasets.
| no_new_dataset | 0.950134 |
1505.04803 | Yong Jae Lee | Yong Jae Lee and Kristen Grauman | Predicting Important Objects for Egocentric Video Summarization | Published in the International Journal of Computer Vision (IJCV),
January 2015 | null | 10.1007/s11263-014-0794-5 | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We present a video summarization approach for egocentric or "wearable" camera
data. Given hours of video, the proposed method produces a compact storyboard
summary of the camera wearer's day. In contrast to traditional keyframe
selection techniques, the resulting summary focuses on the most important
objects and people with which the camera wearer interacts. To accomplish this,
we develop region cues indicative of high-level saliency in egocentric
video---such as the nearness to hands, gaze, and frequency of occurrence---and
learn a regressor to predict the relative importance of any new region based on
these cues. Using these predictions and a simple form of temporal event
detection, our method selects frames for the storyboard that reflect the key
object-driven happenings. We adjust the compactness of the final summary given
either an importance selection criterion or a length budget; for the latter, we
design an efficient dynamic programming solution that accounts for importance,
visual uniqueness, and temporal displacement. Critically, the approach is
neither camera-wearer-specific nor object-specific; that means the learned
importance metric need not be trained for a given user or context, and it can
predict the importance of objects and people that have never been seen
previously. Our results on two egocentric video datasets show the method's
promise relative to existing techniques for saliency and summarization.
| [
{
"version": "v1",
"created": "Mon, 18 May 2015 20:07:20 GMT"
}
] | 2015-05-20T00:00:00 | [
[
"Lee",
"Yong Jae",
""
],
[
"Grauman",
"Kristen",
""
]
] | TITLE: Predicting Important Objects for Egocentric Video Summarization
ABSTRACT: We present a video summarization approach for egocentric or "wearable" camera
data. Given hours of video, the proposed method produces a compact storyboard
summary of the camera wearer's day. In contrast to traditional keyframe
selection techniques, the resulting summary focuses on the most important
objects and people with which the camera wearer interacts. To accomplish this,
we develop region cues indicative of high-level saliency in egocentric
video---such as the nearness to hands, gaze, and frequency of occurrence---and
learn a regressor to predict the relative importance of any new region based on
these cues. Using these predictions and a simple form of temporal event
detection, our method selects frames for the storyboard that reflect the key
object-driven happenings. We adjust the compactness of the final summary given
either an importance selection criterion or a length budget; for the latter, we
design an efficient dynamic programming solution that accounts for importance,
visual uniqueness, and temporal displacement. Critically, the approach is
neither camera-wearer-specific nor object-specific; that means the learned
importance metric need not be trained for a given user or context, and it can
predict the importance of objects and people that have never been seen
previously. Our results on two egocentric video datasets show the method's
promise relative to existing techniques for saliency and summarization.
| no_new_dataset | 0.955068 |
1505.04873 | Lior Talker | Lior Talker and Yael Moses and Ilan Shimshoni | Have a Look at What I See | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We propose a method for guiding a photographer to rotate her/his smartphone
camera to obtain an image that overlaps with another image of the same scene.
The other image is taken by another photographer from a different viewpoint.
Our method is applicable even when the images do not have overlapping fields of
view. Straightforward applications of our method include sharing attention to
regions of interest for social purposes, or adding missing images to improve
structure for motion results. Our solution uses additional images of the scene,
which are often available since many people use their smartphone cameras
regularly. These images may be available online from other photographers who
are present at the scene. Our method avoids 3D scene reconstruction; it relies
instead on a new representation that consists of the spatial orders of the
scene points on two axes, x and y. This representation allows a sequence of
points to be chosen efficiently and projected onto the photographers images,
using epipolar point transfer. Overlaying these epipolar lines on the live
preview of the camera produces a convenient interface to guide the user. The
method was tested on challenging datasets of images and succeeded in guiding a
photographer from one view to a non-overlapping destination view.
| [
{
"version": "v1",
"created": "Tue, 19 May 2015 04:49:46 GMT"
}
] | 2015-05-20T00:00:00 | [
[
"Talker",
"Lior",
""
],
[
"Moses",
"Yael",
""
],
[
"Shimshoni",
"Ilan",
""
]
] | TITLE: Have a Look at What I See
ABSTRACT: We propose a method for guiding a photographer to rotate her/his smartphone
camera to obtain an image that overlaps with another image of the same scene.
The other image is taken by another photographer from a different viewpoint.
Our method is applicable even when the images do not have overlapping fields of
view. Straightforward applications of our method include sharing attention to
regions of interest for social purposes, or adding missing images to improve
structure for motion results. Our solution uses additional images of the scene,
which are often available since many people use their smartphone cameras
regularly. These images may be available online from other photographers who
are present at the scene. Our method avoids 3D scene reconstruction; it relies
instead on a new representation that consists of the spatial orders of the
scene points on two axes, x and y. This representation allows a sequence of
points to be chosen efficiently and projected onto the photographers images,
using epipolar point transfer. Overlaying these epipolar lines on the live
preview of the camera produces a convenient interface to guide the user. The
method was tested on challenging datasets of images and succeeded in guiding a
photographer from one view to a non-overlapping destination view.
| no_new_dataset | 0.942454 |
1505.04880 | Nasser Ghadiri | Amin Beiranvand and Nasser Ghadiri | ADQUEX: Adaptive Processing of Federated Queries over Linked Data based
on Tuple Routing | null | null | null | null | cs.DC cs.DB | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Due to the distribution of linked data across the web, the methods that
process federated queries through a distributed approach are more attractive to
the users and have gained more prosperity. In distributed processing of
federated queries, we need methods and procedures to execute the query in an
optimal manner. Most of the existing methods perform the optimization task
based on some statistical information, whereas the query processor does not
have precise statistical information about their properties, since the data
sources are autonomous. When precise statistics are not available, the
possibility of wrong estimations will highly increase, and may lead to
inefficient execution of the query at runtime. Another problem of the existing
methods is that in the optimization phase, they assume that runtime conditions
of query execution are stable, while the environment in which federated queries
are executed over linked data is dynamic and non-predictable. By considering
these two problems, there is a great potential for exploiting the federated
query processing techniques in an adaptive manner. In this paper, an adaptive
method is proposed for processing federated queries over linked data, based on
the concept of routing the tuples. The proposed method, named ADQUEX, is able
to execute the query effectively without any prior statistical information.
This method can change the query execution plan at runtime so that less
intermediate results are produced. It can also adapt the execution plan to new
situation if unpredicted network latencies arise. Extensive evaluation of our
method by running real queries over well-known linked datasets shows very good
results especially for complex queries.
| [
{
"version": "v1",
"created": "Tue, 19 May 2015 05:51:27 GMT"
}
] | 2015-05-20T00:00:00 | [
[
"Beiranvand",
"Amin",
""
],
[
"Ghadiri",
"Nasser",
""
]
] | TITLE: ADQUEX: Adaptive Processing of Federated Queries over Linked Data based
on Tuple Routing
ABSTRACT: Due to the distribution of linked data across the web, the methods that
process federated queries through a distributed approach are more attractive to
the users and have gained more prosperity. In distributed processing of
federated queries, we need methods and procedures to execute the query in an
optimal manner. Most of the existing methods perform the optimization task
based on some statistical information, whereas the query processor does not
have precise statistical information about their properties, since the data
sources are autonomous. When precise statistics are not available, the
possibility of wrong estimations will highly increase, and may lead to
inefficient execution of the query at runtime. Another problem of the existing
methods is that in the optimization phase, they assume that runtime conditions
of query execution are stable, while the environment in which federated queries
are executed over linked data is dynamic and non-predictable. By considering
these two problems, there is a great potential for exploiting the federated
query processing techniques in an adaptive manner. In this paper, an adaptive
method is proposed for processing federated queries over linked data, based on
the concept of routing the tuples. The proposed method, named ADQUEX, is able
to execute the query effectively without any prior statistical information.
This method can change the query execution plan at runtime so that less
intermediate results are produced. It can also adapt the execution plan to new
situation if unpredicted network latencies arise. Extensive evaluation of our
method by running real queries over well-known linked datasets shows very good
results especially for complex queries.
| no_new_dataset | 0.937897 |
1505.04922 | Lianwen Jin | Weixin Yang, Lianwen Jin, Manfei Liu | Character-level Chinese Writer Identification using Path Signature
Feature, DropStroke and Deep CNN | 5 pages, 4 figures, 2 tables. Manuscript is accepted to appear in
ICDAR 2015 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Most existing online writer-identification systems require that the text
content is supplied in advance and rely on separately designed features and
classifiers. The identifications are based on lines of text, entire paragraphs,
or entire documents; however, these materials are not always available. In this
paper, we introduce a path-signature feature to an end-to-end text-independent
writer-identification system with a deep convolutional neural network (DCNN).
Because deep models require a considerable amount of data to achieve good
performance, we propose a data-augmentation method named DropStroke to enrich
personal handwriting. Experiments were conducted on online handwritten Chinese
characters from the CASIA-OLHWDB1.0 dataset, which consists of 3,866 classes
from 420 writers. For each writer, we only used 200 samples for training and
the remaining 3,666. The results reveal that the path-signature feature is
useful for writer identification, and the proposed DropStroke technique
enhances the generalization and significantly improves performance.
| [
{
"version": "v1",
"created": "Tue, 19 May 2015 09:25:46 GMT"
}
] | 2015-05-20T00:00:00 | [
[
"Yang",
"Weixin",
""
],
[
"Jin",
"Lianwen",
""
],
[
"Liu",
"Manfei",
""
]
] | TITLE: Character-level Chinese Writer Identification using Path Signature
Feature, DropStroke and Deep CNN
ABSTRACT: Most existing online writer-identification systems require that the text
content is supplied in advance and rely on separately designed features and
classifiers. The identifications are based on lines of text, entire paragraphs,
or entire documents; however, these materials are not always available. In this
paper, we introduce a path-signature feature to an end-to-end text-independent
writer-identification system with a deep convolutional neural network (DCNN).
Because deep models require a considerable amount of data to achieve good
performance, we propose a data-augmentation method named DropStroke to enrich
personal handwriting. Experiments were conducted on online handwritten Chinese
characters from the CASIA-OLHWDB1.0 dataset, which consists of 3,866 classes
from 420 writers. For each writer, we only used 200 samples for training and
the remaining 3,666. The results reveal that the path-signature feature is
useful for writer identification, and the proposed DropStroke technique
enhances the generalization and significantly improves performance.
| no_new_dataset | 0.948632 |
1505.04925 | Lianwen Jin | Zhuoyao Zhong, Lianwen Jin, Zecheng Xie | High Performance Offline Handwritten Chinese Character Recognition Using
GoogLeNet and Directional Feature Maps | 5 pages, 4 figures, 2 tables. Manuscript is accepted to appear in
ICDAR 2015 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Just like its great success in solving many computer vision problems, the
convolutional neural networks (CNN) provided new end-to-end approach to
handwritten Chinese character recognition (HCCR) with very promising results in
recent years. However, previous CNNs so far proposed for HCCR were neither deep
enough nor slim enough. We show in this paper that, a deeper architecture can
benefit HCCR a lot to achieve higher performance, meanwhile can be designed
with less parameters. We also show that the traditional feature extraction
methods, such as Gabor or gradient feature maps, are still useful for enhancing
the performance of CNN. We design a streamlined version of GoogLeNet [13],
which was original proposed for image classification in recent years with very
deep architecture, for HCCR (denoted as HCCR-GoogLeNet). The HCCR-GoogLeNet we
used is 19 layers deep but involves with only 7.26 million parameters.
Experiments were conducted using the ICDAR 2013 offline HCCR competition
dataset. It has been shown that with the proper incorporation with traditional
directional feature maps, the proposed single and ensemble HCCR-GoogLeNet
models achieve new state of the art recognition accuracy of 96.35% and 96.74%,
respectively, outperforming previous best result with significant gap.
| [
{
"version": "v1",
"created": "Tue, 19 May 2015 09:32:54 GMT"
}
] | 2015-05-20T00:00:00 | [
[
"Zhong",
"Zhuoyao",
""
],
[
"Jin",
"Lianwen",
""
],
[
"Xie",
"Zecheng",
""
]
] | TITLE: High Performance Offline Handwritten Chinese Character Recognition Using
GoogLeNet and Directional Feature Maps
ABSTRACT: Just like its great success in solving many computer vision problems, the
convolutional neural networks (CNN) provided new end-to-end approach to
handwritten Chinese character recognition (HCCR) with very promising results in
recent years. However, previous CNNs so far proposed for HCCR were neither deep
enough nor slim enough. We show in this paper that, a deeper architecture can
benefit HCCR a lot to achieve higher performance, meanwhile can be designed
with less parameters. We also show that the traditional feature extraction
methods, such as Gabor or gradient feature maps, are still useful for enhancing
the performance of CNN. We design a streamlined version of GoogLeNet [13],
which was original proposed for image classification in recent years with very
deep architecture, for HCCR (denoted as HCCR-GoogLeNet). The HCCR-GoogLeNet we
used is 19 layers deep but involves with only 7.26 million parameters.
Experiments were conducted using the ICDAR 2013 offline HCCR competition
dataset. It has been shown that with the proper incorporation with traditional
directional feature maps, the proposed single and ensemble HCCR-GoogLeNet
models achieve new state of the art recognition accuracy of 96.35% and 96.74%,
respectively, outperforming previous best result with significant gap.
| no_new_dataset | 0.947381 |
1505.05136 | Nicolas Turenne | Nicolas Turenne | A Table-Binning Approach for Visualizing the Past | null | null | null | null | cs.IR cs.DB cs.SI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Large amounts of data are available due to low-cost and high-capacity data
storage equipments. We propose a data exploration/visualization method for
tabular multi-dimensional, time-varying datasets to present selected items in
their global context. The approach is simple and uses a rank-based
visualization and a pattern matching functionality based on temporal profiles.
Ranking categories can be specified in a flexible way and are used instead of
actual values (value reduction into bins) and plotting it over time in an
unevenly quantized representation. Patterns that emerge are matched against a
set of eight predefined temporal profiles. The graphical summarization of
large-scale temporal data is proposed and applicability is tested qualitatively
on about eight data sets and the approach is compared to classic line plots and
SAX representation
| [
{
"version": "v1",
"created": "Mon, 27 Apr 2015 11:15:39 GMT"
}
] | 2015-05-20T00:00:00 | [
[
"Turenne",
"Nicolas",
""
]
] | TITLE: A Table-Binning Approach for Visualizing the Past
ABSTRACT: Large amounts of data are available due to low-cost and high-capacity data
storage equipments. We propose a data exploration/visualization method for
tabular multi-dimensional, time-varying datasets to present selected items in
their global context. The approach is simple and uses a rank-based
visualization and a pattern matching functionality based on temporal profiles.
Ranking categories can be specified in a flexible way and are used instead of
actual values (value reduction into bins) and plotting it over time in an
unevenly quantized representation. Patterns that emerge are matched against a
set of eight predefined temporal profiles. The graphical summarization of
large-scale temporal data is proposed and applicability is tested qualitatively
on about eight data sets and the approach is compared to classic line plots and
SAX representation
| no_new_dataset | 0.948058 |
1005.5227 | Michael Schreiber | Michael Schreiber | Twenty Hirsch index variants and other indicators giving more or less
preference to highly cited papers | 19 pages, including 6 tables and 3 figures accepted for publication
in Annalen der Physik Berlin | Ann. Phys. (Berlin) 522, 536-554 (2010) | 10.1002/andp.201000046 | null | physics.soc-ph cs.DL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The Hirsch index or h-index is widely used to quantify the impact of an
individual's scientific research output, determining the highest number h of a
scientist's papers that received at least h citations. Several variants of the
index have been proposed in order to give more or less preference to highly
cited papers. I analyse the citation records of 26 physicists discussing
various suggestions, in particular A, e, f, g, h(2), h_w, h_T, \hbar, m, {\pi},
R, s, t, w, and maxprod. The total number of all and of all cited publications
as well as the highest and the average number of citations are also compared.
Advantages and disadvantages of these indices and indicators are discussed.
Correlation coefficients are determined quantifying which indices and
indicators yield similar and which yield more deviating rankings of the 26
datasets. For 6 datasets the determination of the indices and indicators is
visualized.
| [
{
"version": "v1",
"created": "Fri, 28 May 2010 07:08:09 GMT"
}
] | 2015-05-19T00:00:00 | [
[
"Schreiber",
"Michael",
""
]
] | TITLE: Twenty Hirsch index variants and other indicators giving more or less
preference to highly cited papers
ABSTRACT: The Hirsch index or h-index is widely used to quantify the impact of an
individual's scientific research output, determining the highest number h of a
scientist's papers that received at least h citations. Several variants of the
index have been proposed in order to give more or less preference to highly
cited papers. I analyse the citation records of 26 physicists discussing
various suggestions, in particular A, e, f, g, h(2), h_w, h_T, \hbar, m, {\pi},
R, s, t, w, and maxprod. The total number of all and of all cited publications
as well as the highest and the average number of citations are also compared.
Advantages and disadvantages of these indices and indicators are discussed.
Correlation coefficients are determined quantifying which indices and
indicators yield similar and which yield more deviating rankings of the 26
datasets. For 6 datasets the determination of the indices and indicators is
visualized.
| no_new_dataset | 0.952042 |
1006.4828 | Vamsi Kundeti | Vamsi Kundeti and Sanguthevar Rajasekaran and Hieu Dinh | An Efficient Algorithm For Chinese Postman Walk on Bi-directed de Bruijn
Graphs | null | null | 10.1007/978-3-642-17458-2_16 | null | cs.DS | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Sequence assembly from short reads is an important problem in biology. It is
known that solving the sequence assembly problem exactly on a bi-directed de
Bruijn graph or a string graph is intractable. However finding a Shortest
Double stranded DNA string (SDDNA) containing all the k-long words in the reads
seems to be a good heuristic to get close to the original genome. This problem
is equivalent to finding a cyclic Chinese Postman (CP) walk on the underlying
un-weighted bi-directed de Bruijn graph built from the reads. The Chinese
Postman walk Problem (CPP) is solved by reducing it to a general bi-directed
flow on this graph which runs in O(|E|2 log2(|V |)) time. In this paper we show
that the cyclic CPP on bi-directed graphs can be solved without reducing it to
bi-directed flow. We present a ?(p(|V | + |E|) log(|V |) + (dmaxp)3) time
algorithm to solve the cyclic CPP on a weighted bi-directed de Bruijn graph,
where p = max{|{v|din(v) - dout(v) > 0}|, |{v|din(v) - dout(v) < 0}|} and dmax
= max{|din(v) - dout(v)}. Our algorithm performs asymptotically better than the
bidirected flow algorithm when the number of imbalanced nodes p is much less
than the nodes in the bi-directed graph. From our experimental results on
various datasets, we have noticed that the value of p/|V | lies between 0.08%
and 0.13% with 95% probability.
| [
{
"version": "v1",
"created": "Thu, 24 Jun 2010 16:35:47 GMT"
}
] | 2015-05-19T00:00:00 | [
[
"Kundeti",
"Vamsi",
""
],
[
"Rajasekaran",
"Sanguthevar",
""
],
[
"Dinh",
"Hieu",
""
]
] | TITLE: An Efficient Algorithm For Chinese Postman Walk on Bi-directed de Bruijn
Graphs
ABSTRACT: Sequence assembly from short reads is an important problem in biology. It is
known that solving the sequence assembly problem exactly on a bi-directed de
Bruijn graph or a string graph is intractable. However finding a Shortest
Double stranded DNA string (SDDNA) containing all the k-long words in the reads
seems to be a good heuristic to get close to the original genome. This problem
is equivalent to finding a cyclic Chinese Postman (CP) walk on the underlying
un-weighted bi-directed de Bruijn graph built from the reads. The Chinese
Postman walk Problem (CPP) is solved by reducing it to a general bi-directed
flow on this graph which runs in O(|E|2 log2(|V |)) time. In this paper we show
that the cyclic CPP on bi-directed graphs can be solved without reducing it to
bi-directed flow. We present a ?(p(|V | + |E|) log(|V |) + (dmaxp)3) time
algorithm to solve the cyclic CPP on a weighted bi-directed de Bruijn graph,
where p = max{|{v|din(v) - dout(v) > 0}|, |{v|din(v) - dout(v) < 0}|} and dmax
= max{|din(v) - dout(v)}. Our algorithm performs asymptotically better than the
bidirected flow algorithm when the number of imbalanced nodes p is much less
than the nodes in the bi-directed graph. From our experimental results on
various datasets, we have noticed that the value of p/|V | lies between 0.08%
and 0.13% with 95% probability.
| no_new_dataset | 0.952838 |
1008.3907 | Michael Murphy | J. K. Webb, J. A. King, M. T. Murphy, V. V. Flambaum, R. F. Carswell,
M. B. Bainbridge | Indications of a spatial variation of the fine structure constant | 5 pages, 5 figures, published on 31st October 2011 in Physical Review
Letters | Phys. Rev. Lett., 107, 191101, 2011 | 10.1103/PhysRevLett.107.191101 | null | astro-ph.CO gr-qc hep-ph hep-th nucl-th physics.atom-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We previously reported Keck telescope observations suggesting a smaller value
of the fine structure constant, alpha, at high redshift. New Very Large
Telescope (VLT) data, probing a different direction in the universe, shows an
inverse evolution; alpha increases at high redshift. Although the pattern could
be due to as yet undetected systematic effects, with the systematics as
presently understood the combined dataset fits a spatial dipole, significant at
the 4.2-sigma level, in the direction right ascension 17.5 +/- 0.9 hours,
declination -58 +/- 9 degrees. The independent VLT and Keck samples give
consistent dipole directions and amplitudes, as do high and low redshift
samples. A search for systematics, using observations duplicated at both
telescopes, reveals none so far which emulate this result.
| [
{
"version": "v1",
"created": "Mon, 23 Aug 2010 20:00:12 GMT"
},
{
"version": "v2",
"created": "Tue, 1 Nov 2011 03:06:27 GMT"
}
] | 2015-05-19T00:00:00 | [
[
"Webb",
"J. K.",
""
],
[
"King",
"J. A.",
""
],
[
"Murphy",
"M. T.",
""
],
[
"Flambaum",
"V. V.",
""
],
[
"Carswell",
"R. F.",
""
],
[
"Bainbridge",
"M. B.",
""
]
] | TITLE: Indications of a spatial variation of the fine structure constant
ABSTRACT: We previously reported Keck telescope observations suggesting a smaller value
of the fine structure constant, alpha, at high redshift. New Very Large
Telescope (VLT) data, probing a different direction in the universe, shows an
inverse evolution; alpha increases at high redshift. Although the pattern could
be due to as yet undetected systematic effects, with the systematics as
presently understood the combined dataset fits a spatial dipole, significant at
the 4.2-sigma level, in the direction right ascension 17.5 +/- 0.9 hours,
declination -58 +/- 9 degrees. The independent VLT and Keck samples give
consistent dipole directions and amplitudes, as do high and low redshift
samples. A search for systematics, using observations duplicated at both
telescopes, reveals none so far which emulate this result.
| no_new_dataset | 0.930268 |
1008.3982 | Peng Wang | Peng Wang, Ting Lei, Chi Ho Yeung and Bing-hong Wang | Heterogenous Human Dynamics in Intra and Inter-day Time Scale | null | null | 10.1209/0295-5075/94/18005 | null | physics.soc-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper, we study two large data sets containing the information of two
different human behaviors: blog-posting and wiki-revising. In both cases, the
interevent time distributions decay as power-laws at both individual and
population level. As different from previous studies, we put emphasis on time
scales and obtain heterogeneous decay exponents in intra- and inter-day range
for the same dataset. Moreover, we observe opposite trend of exponents in
relation to individual $Activity$. Further investigations show that the
presence of intra-day activities mask the correlation between consecutive
inter-day activities and lead to an underestimate of $Memory$, which explain
the contradicting results in recent empirical studies. Removal of data in
intra-day range reveals the high values of $Memory$ and lead us to convergent
results between wiki-revising and blog-posting.
| [
{
"version": "v1",
"created": "Tue, 24 Aug 2010 08:18:47 GMT"
}
] | 2015-05-19T00:00:00 | [
[
"Wang",
"Peng",
""
],
[
"Lei",
"Ting",
""
],
[
"Yeung",
"Chi Ho",
""
],
[
"Wang",
"Bing-hong",
""
]
] | TITLE: Heterogenous Human Dynamics in Intra and Inter-day Time Scale
ABSTRACT: In this paper, we study two large data sets containing the information of two
different human behaviors: blog-posting and wiki-revising. In both cases, the
interevent time distributions decay as power-laws at both individual and
population level. As different from previous studies, we put emphasis on time
scales and obtain heterogeneous decay exponents in intra- and inter-day range
for the same dataset. Moreover, we observe opposite trend of exponents in
relation to individual $Activity$. Further investigations show that the
presence of intra-day activities mask the correlation between consecutive
inter-day activities and lead to an underestimate of $Memory$, which explain
the contradicting results in recent empirical studies. Removal of data in
intra-day range reveals the high values of $Memory$ and lead us to convergent
results between wiki-revising and blog-posting.
| no_new_dataset | 0.94079 |
1404.1653 | Zi-Ke Zhang Dr. | Lu Yu, Chuang Liu, Zi-Ke Zhang | Multi-Linear Interactive Matrix Factorization | null | null | null | null | cs.IR | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Recommender systems, which can significantly help users find their interested
items from the information era, has attracted an increasing attention from both
the scientific and application society. One of the widest applied
recommendation methods is the Matrix Factorization (MF). However, most of MF
based approaches focus on the user-item rating matrix, but ignoring the
ingredients which may have significant influence on users' preferences on
items. In this paper, we propose a multi-linear interactive MF algorithm
(MLIMF) to model the interactions between the users and each event associated
with their final decisions. Our model considers not only the user-item rating
information but also the pairwise interactions based on some empirically
supported factors. In addition, we compared the proposed model with three
typical other methods: user-based collaborative filtering (UCF), item-based
collaborative filtering (ICF) and regularized MF (RMF). Experimental results on
two real-world datasets, \emph{MovieLens} 1M and \emph{MovieLens} 100k, show
that our method performs much better than other three methods in the accuracy
of recommendation. This work may shed some light on the in-depth understanding
of modeling user online behaviors and the consequent decisions.
| [
{
"version": "v1",
"created": "Mon, 7 Apr 2014 04:39:14 GMT"
},
{
"version": "v2",
"created": "Mon, 18 May 2015 06:05:02 GMT"
}
] | 2015-05-19T00:00:00 | [
[
"Yu",
"Lu",
""
],
[
"Liu",
"Chuang",
""
],
[
"Zhang",
"Zi-Ke",
""
]
] | TITLE: Multi-Linear Interactive Matrix Factorization
ABSTRACT: Recommender systems, which can significantly help users find their interested
items from the information era, has attracted an increasing attention from both
the scientific and application society. One of the widest applied
recommendation methods is the Matrix Factorization (MF). However, most of MF
based approaches focus on the user-item rating matrix, but ignoring the
ingredients which may have significant influence on users' preferences on
items. In this paper, we propose a multi-linear interactive MF algorithm
(MLIMF) to model the interactions between the users and each event associated
with their final decisions. Our model considers not only the user-item rating
information but also the pairwise interactions based on some empirically
supported factors. In addition, we compared the proposed model with three
typical other methods: user-based collaborative filtering (UCF), item-based
collaborative filtering (ICF) and regularized MF (RMF). Experimental results on
two real-world datasets, \emph{MovieLens} 1M and \emph{MovieLens} 100k, show
that our method performs much better than other three methods in the accuracy
of recommendation. This work may shed some light on the in-depth understanding
of modeling user online behaviors and the consequent decisions.
| no_new_dataset | 0.946941 |
1407.2433 | Peter Foster | Peter Foster, Simon Dixon, Anssi Klapuri | Identifying Cover Songs Using Information-Theoretic Measures of
Similarity | 13 pages, 5 figures, 4 tables. v3: Accepted version | IEEE/ACM Transactions on Audio, Speech, and Language Processing,
vol. 23 no. 6, pp. 993-1005, 2015 | 10.1109/TASLP.2015.2416655 | null | cs.IR cs.LG stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper investigates methods for quantifying similarity between audio
signals, specifically for the task of of cover song detection. We consider an
information-theoretic approach, where we compute pairwise measures of
predictability between time series. We compare discrete-valued approaches
operating on quantised audio features, to continuous-valued approaches. In the
discrete case, we propose a method for computing the normalised compression
distance, where we account for correlation between time series. In the
continuous case, we propose to compute information-based measures of similarity
as statistics of the prediction error between time series. We evaluate our
methods on two cover song identification tasks using a data set comprised of
300 Jazz standards and using the Million Song Dataset. For both datasets, we
observe that continuous-valued approaches outperform discrete-valued
approaches. We consider approaches to estimating the normalised compression
distance (NCD) based on string compression and prediction, where we observe
that our proposed normalised compression distance with alignment (NCDA)
improves average performance over NCD, for sequential compression algorithms.
Finally, we demonstrate that continuous-valued distances may be combined to
improve performance with respect to baseline approaches. Using a large-scale
filter-and-refine approach, we demonstrate state-of-the-art performance for
cover song identification using the Million Song Dataset.
| [
{
"version": "v1",
"created": "Wed, 9 Jul 2014 11:04:15 GMT"
},
{
"version": "v2",
"created": "Thu, 10 Jul 2014 00:53:47 GMT"
},
{
"version": "v3",
"created": "Sun, 17 May 2015 15:53:43 GMT"
}
] | 2015-05-19T00:00:00 | [
[
"Foster",
"Peter",
""
],
[
"Dixon",
"Simon",
""
],
[
"Klapuri",
"Anssi",
""
]
] | TITLE: Identifying Cover Songs Using Information-Theoretic Measures of
Similarity
ABSTRACT: This paper investigates methods for quantifying similarity between audio
signals, specifically for the task of of cover song detection. We consider an
information-theoretic approach, where we compute pairwise measures of
predictability between time series. We compare discrete-valued approaches
operating on quantised audio features, to continuous-valued approaches. In the
discrete case, we propose a method for computing the normalised compression
distance, where we account for correlation between time series. In the
continuous case, we propose to compute information-based measures of similarity
as statistics of the prediction error between time series. We evaluate our
methods on two cover song identification tasks using a data set comprised of
300 Jazz standards and using the Million Song Dataset. For both datasets, we
observe that continuous-valued approaches outperform discrete-valued
approaches. We consider approaches to estimating the normalised compression
distance (NCD) based on string compression and prediction, where we observe
that our proposed normalised compression distance with alignment (NCDA)
improves average performance over NCD, for sequential compression algorithms.
Finally, we demonstrate that continuous-valued distances may be combined to
improve performance with respect to baseline approaches. Using a large-scale
filter-and-refine approach, we demonstrate state-of-the-art performance for
cover song identification using the Million Song Dataset.
| no_new_dataset | 0.939526 |
1501.07645 | Sachin Talathi | Sachin S. Talathi | Hyper-parameter optimization of Deep Convolutional Networks for object
recognition | 4 pages, 1 figure, 3 tables, Submitted to ICIP 2015 | null | null | null | cs.CV cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Recently sequential model based optimization (SMBO) has emerged as a
promising hyper-parameter optimization strategy in machine learning. In this
work, we investigate SMBO to identify architecture hyper-parameters of deep
convolution networks (DCNs) object recognition. We propose a simple SMBO
strategy that starts from a set of random initial DCN architectures to generate
new architectures, which on training perform well on a given dataset. Using the
proposed SMBO strategy we are able to identify a number of DCN architectures
that produce results that are comparable to state-of-the-art results on object
recognition benchmarks.
| [
{
"version": "v1",
"created": "Fri, 30 Jan 2015 02:08:51 GMT"
},
{
"version": "v2",
"created": "Sun, 17 May 2015 03:32:22 GMT"
}
] | 2015-05-19T00:00:00 | [
[
"Talathi",
"Sachin S.",
""
]
] | TITLE: Hyper-parameter optimization of Deep Convolutional Networks for object
recognition
ABSTRACT: Recently sequential model based optimization (SMBO) has emerged as a
promising hyper-parameter optimization strategy in machine learning. In this
work, we investigate SMBO to identify architecture hyper-parameters of deep
convolution networks (DCNs) object recognition. We propose a simple SMBO
strategy that starts from a set of random initial DCN architectures to generate
new architectures, which on training perform well on a given dataset. Using the
proposed SMBO strategy we are able to identify a number of DCN architectures
that produce results that are comparable to state-of-the-art results on object
recognition benchmarks.
| no_new_dataset | 0.954052 |
1504.02531 | Zhimin Gao | Zhimin Gao, Lei Wang, Luping Zhou, Jianjia Zhang | HEp-2 Cell Image Classification with Deep Convolutional Neural Networks | 32 pages | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Efficient Human Epithelial-2 (HEp-2) cell image classification can facilitate
the diagnosis of many autoimmune diseases. This paper presents an automatic
framework for this classification task, by utilizing the deep convolutional
neural networks (CNNs) which have recently attracted intensive attention in
visual recognition. This paper elaborates the important components of this
framework, discusses multiple key factors that impact the efficiency of
training a deep CNN, and systematically compares this framework with the
well-established image classification models in the literature. Experiments on
benchmark datasets show that i) the proposed framework can effectively
outperform existing models by properly applying data augmentation; ii) our
CNN-based framework demonstrates excellent adaptability across different
datasets, which is highly desirable for classification under varying laboratory
settings. Our system is ranked high in the cell image classification
competition hosted by ICPR 2014.
| [
{
"version": "v1",
"created": "Fri, 10 Apr 2015 01:58:17 GMT"
},
{
"version": "v2",
"created": "Mon, 18 May 2015 01:12:12 GMT"
}
] | 2015-05-19T00:00:00 | [
[
"Gao",
"Zhimin",
""
],
[
"Wang",
"Lei",
""
],
[
"Zhou",
"Luping",
""
],
[
"Zhang",
"Jianjia",
""
]
] | TITLE: HEp-2 Cell Image Classification with Deep Convolutional Neural Networks
ABSTRACT: Efficient Human Epithelial-2 (HEp-2) cell image classification can facilitate
the diagnosis of many autoimmune diseases. This paper presents an automatic
framework for this classification task, by utilizing the deep convolutional
neural networks (CNNs) which have recently attracted intensive attention in
visual recognition. This paper elaborates the important components of this
framework, discusses multiple key factors that impact the efficiency of
training a deep CNN, and systematically compares this framework with the
well-established image classification models in the literature. Experiments on
benchmark datasets show that i) the proposed framework can effectively
outperform existing models by properly applying data augmentation; ii) our
CNN-based framework demonstrates excellent adaptability across different
datasets, which is highly desirable for classification under varying laboratory
settings. Our system is ranked high in the cell image classification
competition hosted by ICPR 2014.
| no_new_dataset | 0.952397 |
1504.06236 | Arman Shokrollahi | Amir Sheikhahmadi, Mohammad A. Nematbakhsh, and Arman Shokrollahi | Improving detection of influential nodes in complex networks | 17 pages, 8 figures, 5 tables . . . accepted for publication in
Physica A: Statistical Mechanics and its Applications | null | 10.1016/j.physa.2015.04.035 | null | cs.SI physics.soc-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Recently an increasing amount of research is devoted to the question of how
the most influential nodes (seeds) can be found effectively in a complex
network. There are a number of measures proposed for this purpose, for
instance, high-degree centrality measure reflects the importance of the network
topology and has a reasonable runtime performance to find a set of nodes with
highest degree, but they do not have a satisfactory dissemination potentiality
in the network due to having many common neighbors ($\mbox{CN}^{(1)}$) and
common neighbors of neighbors ($\mbox{CN}^{(2)}$). This flaw holds in other
measures as well. In this paper, we compare high-degree centrality measure with
other well-known measures using ten datasets in order to find a proportion for
the common seeds in the seed sets obtained by them. We, thereof, propose an
improved high-degree centrality measure (named DegreeDistance) and improve it
to enhance accuracy in two phases, FIDD and SIDD, by putting a threshold on the
number of common neighbors of already-selected seed nodes and a non-seed node
which is under investigation to be selected as a seed as well as considering
the influence score of seed nodes directly or through their common neighbors
over the non-seed node. To evaluate the accuracy and runtime performance of
DegreeDistance, FIDD, and SIDD, they are applied to eight large-scale networks
and it finally turns out that SIDD dramatically outperforms other well-known
measures and evinces comparatively more accurate performance in identifying the
most influential nodes.
| [
{
"version": "v1",
"created": "Thu, 23 Apr 2015 16:01:35 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Apr 2015 08:10:49 GMT"
}
] | 2015-05-19T00:00:00 | [
[
"Sheikhahmadi",
"Amir",
""
],
[
"Nematbakhsh",
"Mohammad A.",
""
],
[
"Shokrollahi",
"Arman",
""
]
] | TITLE: Improving detection of influential nodes in complex networks
ABSTRACT: Recently an increasing amount of research is devoted to the question of how
the most influential nodes (seeds) can be found effectively in a complex
network. There are a number of measures proposed for this purpose, for
instance, high-degree centrality measure reflects the importance of the network
topology and has a reasonable runtime performance to find a set of nodes with
highest degree, but they do not have a satisfactory dissemination potentiality
in the network due to having many common neighbors ($\mbox{CN}^{(1)}$) and
common neighbors of neighbors ($\mbox{CN}^{(2)}$). This flaw holds in other
measures as well. In this paper, we compare high-degree centrality measure with
other well-known measures using ten datasets in order to find a proportion for
the common seeds in the seed sets obtained by them. We, thereof, propose an
improved high-degree centrality measure (named DegreeDistance) and improve it
to enhance accuracy in two phases, FIDD and SIDD, by putting a threshold on the
number of common neighbors of already-selected seed nodes and a non-seed node
which is under investigation to be selected as a seed as well as considering
the influence score of seed nodes directly or through their common neighbors
over the non-seed node. To evaluate the accuracy and runtime performance of
DegreeDistance, FIDD, and SIDD, they are applied to eight large-scale networks
and it finally turns out that SIDD dramatically outperforms other well-known
measures and evinces comparatively more accurate performance in identifying the
most influential nodes.
| no_new_dataset | 0.951997 |
1505.04243 | Paul Grigas | Robert M. Freund, Paul Grigas, Rahul Mazumder | A New Perspective on Boosting in Linear Regression via Subgradient
Optimization and Relatives | null | null | null | null | math.ST cs.LG math.OC stat.ML stat.TH | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper we analyze boosting algorithms in linear regression from a new
perspective: that of modern first-order methods in convex optimization. We show
that classic boosting algorithms in linear regression, namely the incremental
forward stagewise algorithm (FS$_\varepsilon$) and least squares boosting
(LS-Boost($\varepsilon$)), can be viewed as subgradient descent to minimize the
loss function defined as the maximum absolute correlation between the features
and residuals. We also propose a modification of FS$_\varepsilon$ that yields
an algorithm for the Lasso, and that may be easily extended to an algorithm
that computes the Lasso path for different values of the regularization
parameter. Furthermore, we show that these new algorithms for the Lasso may
also be interpreted as the same master algorithm (subgradient descent), applied
to a regularized version of the maximum absolute correlation loss function. We
derive novel, comprehensive computational guarantees for several boosting
algorithms in linear regression (including LS-Boost($\varepsilon$) and
FS$_\varepsilon$) by using techniques of modern first-order methods in convex
optimization. Our computational guarantees inform us about the statistical
properties of boosting algorithms. In particular they provide, for the first
time, a precise theoretical description of the amount of data-fidelity and
regularization imparted by running a boosting algorithm with a prespecified
learning rate for a fixed but arbitrary number of iterations, for any dataset.
| [
{
"version": "v1",
"created": "Sat, 16 May 2015 04:23:08 GMT"
}
] | 2015-05-19T00:00:00 | [
[
"Freund",
"Robert M.",
""
],
[
"Grigas",
"Paul",
""
],
[
"Mazumder",
"Rahul",
""
]
] | TITLE: A New Perspective on Boosting in Linear Regression via Subgradient
Optimization and Relatives
ABSTRACT: In this paper we analyze boosting algorithms in linear regression from a new
perspective: that of modern first-order methods in convex optimization. We show
that classic boosting algorithms in linear regression, namely the incremental
forward stagewise algorithm (FS$_\varepsilon$) and least squares boosting
(LS-Boost($\varepsilon$)), can be viewed as subgradient descent to minimize the
loss function defined as the maximum absolute correlation between the features
and residuals. We also propose a modification of FS$_\varepsilon$ that yields
an algorithm for the Lasso, and that may be easily extended to an algorithm
that computes the Lasso path for different values of the regularization
parameter. Furthermore, we show that these new algorithms for the Lasso may
also be interpreted as the same master algorithm (subgradient descent), applied
to a regularized version of the maximum absolute correlation loss function. We
derive novel, comprehensive computational guarantees for several boosting
algorithms in linear regression (including LS-Boost($\varepsilon$) and
FS$_\varepsilon$) by using techniques of modern first-order methods in convex
optimization. Our computational guarantees inform us about the statistical
properties of boosting algorithms. In particular they provide, for the first
time, a precise theoretical description of the amount of data-fidelity and
regularization imparted by running a boosting algorithm with a prespecified
learning rate for a fixed but arbitrary number of iterations, for any dataset.
| no_new_dataset | 0.944536 |
1505.04286 | Harry Commin PhD | Harry Commin | Robust Real-time Extraction of Fiducial Facial Feature Points using
Haar-like Features | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper, we explore methods of robustly extracting fiducial facial
feature points - an important process for numerous facial image processing
tasks. We consider various methods to first detect face, then facial features
and finally salient facial feature points. Colour-based models are analysed and
their overall unsuitability for this task is summarised. The bulk of the report
is then dedicated to proposing a learning-based method centred on the
Viola-Jones algorithm. The specific difficulties and considerations relating to
feature point detection are laid out in this context and a novel approach is
established to address these issues. On a sequence of clear and unobstructed
face images, our proposed system achieves average detection rates of over 90%.
Then, using a more varied sample dataset, we identify some possible areas for
future development of our system.
| [
{
"version": "v1",
"created": "Sat, 16 May 2015 15:41:21 GMT"
}
] | 2015-05-19T00:00:00 | [
[
"Commin",
"Harry",
""
]
] | TITLE: Robust Real-time Extraction of Fiducial Facial Feature Points using
Haar-like Features
ABSTRACT: In this paper, we explore methods of robustly extracting fiducial facial
feature points - an important process for numerous facial image processing
tasks. We consider various methods to first detect face, then facial features
and finally salient facial feature points. Colour-based models are analysed and
their overall unsuitability for this task is summarised. The bulk of the report
is then dedicated to proposing a learning-based method centred on the
Viola-Jones algorithm. The specific difficulties and considerations relating to
feature point detection are laid out in this context and a novel approach is
established to address these issues. On a sequence of clear and unobstructed
face images, our proposed system achieves average detection rates of over 90%.
Then, using a more varied sample dataset, we identify some possible areas for
future development of our system.
| no_new_dataset | 0.945651 |
1505.04366 | Seunghoon Hong | Hyeonwoo Noh, Seunghoon Hong, Bohyung Han | Learning Deconvolution Network for Semantic Segmentation | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We propose a novel semantic segmentation algorithm by learning a
deconvolution network. We learn the network on top of the convolutional layers
adopted from VGG 16-layer net. The deconvolution network is composed of
deconvolution and unpooling layers, which identify pixel-wise class labels and
predict segmentation masks. We apply the trained network to each proposal in an
input image, and construct the final semantic segmentation map by combining the
results from all proposals in a simple manner. The proposed algorithm mitigates
the limitations of the existing methods based on fully convolutional networks
by integrating deep deconvolution network and proposal-wise prediction; our
segmentation method typically identifies detailed structures and handles
objects in multiple scales naturally. Our network demonstrates outstanding
performance in PASCAL VOC 2012 dataset, and we achieve the best accuracy
(72.5%) among the methods trained with no external data through ensemble with
the fully convolutional network.
| [
{
"version": "v1",
"created": "Sun, 17 May 2015 07:33:28 GMT"
}
] | 2015-05-19T00:00:00 | [
[
"Noh",
"Hyeonwoo",
""
],
[
"Hong",
"Seunghoon",
""
],
[
"Han",
"Bohyung",
""
]
] | TITLE: Learning Deconvolution Network for Semantic Segmentation
ABSTRACT: We propose a novel semantic segmentation algorithm by learning a
deconvolution network. We learn the network on top of the convolutional layers
adopted from VGG 16-layer net. The deconvolution network is composed of
deconvolution and unpooling layers, which identify pixel-wise class labels and
predict segmentation masks. We apply the trained network to each proposal in an
input image, and construct the final semantic segmentation map by combining the
results from all proposals in a simple manner. The proposed algorithm mitigates
the limitations of the existing methods based on fully convolutional networks
by integrating deep deconvolution network and proposal-wise prediction; our
segmentation method typically identifies detailed structures and handles
objects in multiple scales naturally. Our network demonstrates outstanding
performance in PASCAL VOC 2012 dataset, and we achieve the best accuracy
(72.5%) among the methods trained with no external data through ensemble with
the fully convolutional network.
| no_new_dataset | 0.951459 |
1505.04427 | Zhenzhong Lan | Zhenzhong Lan, Dezhong Yao, Ming Lin, Shoou-I Yu, Alexander Hauptmann | The Best of Both Worlds: Combining Data-independent and Data-driven
Approaches for Action Recognition | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Motivated by the success of data-driven convolutional neural networks (CNNs)
in object recognition on static images, researchers are working hard towards
developing CNN equivalents for learning video features. However, learning video
features globally has proven to be quite a challenge due to its high
dimensionality, the lack of labelled data and the difficulty in processing
large-scale video data. Therefore, we propose to leverage effective techniques
from both data-driven and data-independent approaches to improve action
recognition system.
Our contribution is three-fold. First, we propose a two-stream Stacked
Convolutional Independent Subspace Analysis (ConvISA) architecture to show that
unsupervised learning methods can significantly boost the performance of
traditional local features extracted from data-independent models. Second, we
demonstrate that by learning on video volumes detected by Improved Dense
Trajectory (IDT), we can seamlessly combine our novel local descriptors with
hand-crafted descriptors. Thus we can utilize available feature enhancing
techniques developed for hand-crafted descriptors. Finally, similar to
multi-class classification framework in CNNs, we propose a training-free
re-ranking technique that exploits the relationship among action classes to
improve the overall performance. Our experimental results on four benchmark
action recognition datasets show significantly improved performance.
| [
{
"version": "v1",
"created": "Sun, 17 May 2015 17:54:38 GMT"
}
] | 2015-05-19T00:00:00 | [
[
"Lan",
"Zhenzhong",
""
],
[
"Yao",
"Dezhong",
""
],
[
"Lin",
"Ming",
""
],
[
"Yu",
"Shoou-I",
""
],
[
"Hauptmann",
"Alexander",
""
]
] | TITLE: The Best of Both Worlds: Combining Data-independent and Data-driven
Approaches for Action Recognition
ABSTRACT: Motivated by the success of data-driven convolutional neural networks (CNNs)
in object recognition on static images, researchers are working hard towards
developing CNN equivalents for learning video features. However, learning video
features globally has proven to be quite a challenge due to its high
dimensionality, the lack of labelled data and the difficulty in processing
large-scale video data. Therefore, we propose to leverage effective techniques
from both data-driven and data-independent approaches to improve action
recognition system.
Our contribution is three-fold. First, we propose a two-stream Stacked
Convolutional Independent Subspace Analysis (ConvISA) architecture to show that
unsupervised learning methods can significantly boost the performance of
traditional local features extracted from data-independent models. Second, we
demonstrate that by learning on video volumes detected by Improved Dense
Trajectory (IDT), we can seamlessly combine our novel local descriptors with
hand-crafted descriptors. Thus we can utilize available feature enhancing
techniques developed for hand-crafted descriptors. Finally, similar to
multi-class classification framework in CNNs, we propose a training-free
re-ranking technique that exploits the relationship among action classes to
improve the overall performance. Our experimental results on four benchmark
action recognition datasets show significantly improved performance.
| no_new_dataset | 0.948728 |
1505.04474 | Saurabh Gupta | Saurabh Gupta and Jitendra Malik | Visual Semantic Role Labeling | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper we introduce the problem of Visual Semantic Role Labeling:
given an image we want to detect people doing actions and localize the objects
of interaction. Classical approaches to action recognition either study the
task of action classification at the image or video clip level or at best
produce a bounding box around the person doing the action. We believe such an
output is inadequate and a complete understanding can only come when we are
able to associate objects in the scene to the different semantic roles of the
action. To enable progress towards this goal, we annotate a dataset of 16K
people instances in 10K images with actions they are doing and associate
objects in the scene with different semantic roles for each action. Finally, we
provide a set of baseline algorithms for this task and analyze error modes
providing directions for future work.
| [
{
"version": "v1",
"created": "Sun, 17 May 2015 23:21:35 GMT"
}
] | 2015-05-19T00:00:00 | [
[
"Gupta",
"Saurabh",
""
],
[
"Malik",
"Jitendra",
""
]
] | TITLE: Visual Semantic Role Labeling
ABSTRACT: In this paper we introduce the problem of Visual Semantic Role Labeling:
given an image we want to detect people doing actions and localize the objects
of interaction. Classical approaches to action recognition either study the
task of action classification at the image or video clip level or at best
produce a bounding box around the person doing the action. We believe such an
output is inadequate and a complete understanding can only come when we are
able to associate objects in the scene to the different semantic roles of the
action. To enable progress towards this goal, we annotate a dataset of 16K
people instances in 10K images with actions they are doing and associate
objects in the scene with different semantic roles for each action. Finally, we
provide a set of baseline algorithms for this task and analyze error modes
providing directions for future work.
| no_new_dataset | 0.927166 |
1505.04636 | Alex Smola J | Mu Li, Dave G. Andersen, Alexander J. Smola | Graph Partitioning via Parallel Submodular Approximation to Accelerate
Distributed Machine Learning | null | null | null | null | cs.DC cs.AI cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Distributed computing excels at processing large scale data, but the
communication cost for synchronizing the shared parameters may slow down the
overall performance. Fortunately, the interactions between parameter and data
in many problems are sparse, which admits efficient partition in order to
reduce the communication overhead.
In this paper, we formulate data placement as a graph partitioning problem.
We propose a distributed partitioning algorithm. We give both theoretical
guarantees and a highly efficient implementation. We also provide a highly
efficient implementation of the algorithm and demonstrate its promising results
on both text datasets and social networks. We show that the proposed algorithm
leads to 1.6x speedup of a state-of-the-start distributed machine learning
system by eliminating 90\% of the network communication.
| [
{
"version": "v1",
"created": "Mon, 18 May 2015 13:43:46 GMT"
}
] | 2015-05-19T00:00:00 | [
[
"Li",
"Mu",
""
],
[
"Andersen",
"Dave G.",
""
],
[
"Smola",
"Alexander J.",
""
]
] | TITLE: Graph Partitioning via Parallel Submodular Approximation to Accelerate
Distributed Machine Learning
ABSTRACT: Distributed computing excels at processing large scale data, but the
communication cost for synchronizing the shared parameters may slow down the
overall performance. Fortunately, the interactions between parameter and data
in many problems are sparse, which admits efficient partition in order to
reduce the communication overhead.
In this paper, we formulate data placement as a graph partitioning problem.
We propose a distributed partitioning algorithm. We give both theoretical
guarantees and a highly efficient implementation. We also provide a highly
efficient implementation of the algorithm and demonstrate its promising results
on both text datasets and social networks. We show that the proposed algorithm
leads to 1.6x speedup of a state-of-the-start distributed machine learning
system by eliminating 90\% of the network communication.
| no_new_dataset | 0.945601 |
1002.0157 | Valerio Lucarini | Valerio Lucarini, Klaus Fraedrich, Francesco Ragone | New Results on the Thermodynamical Properties of the Climate System | 56 pages, 6 Figure, 4 Tables. Final version with additional
theoretical discussion, improved analysis of IPCC models. Accepted in J.
Atmos. Sci | null | 10.1175/2011JAS3713.1 | null | physics.ao-ph astro-ph.EP cond-mat.stat-mech physics.class-ph physics.comp-ph physics.flu-dyn physics.geo-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper we exploit two equivalent formulations of the average rate of
material entropy production in a planetary system to propose an approximate
splitting between contributions due to vertical and eminently horizontal
processes. Our approach is based only upon 2D radiative fields at the surface
and at the top of atmosphere of a general planetary body. Using 2D fields at
the top of atmosphere alone, we derive lower bounds to the rate of material
entropy production and to the intensity of the Lorenz energy cycle. By
introducing a measure of the efficiency of the planetary system with respect to
horizontal thermodynamical processes, we provide insight on a previous
intuition on the possibility of defining a baroclinic heat engine extracting
work from the meridional heat flux. The approximate formula of the material
entropy production is verified and used for studying the global thermodynamic
properties of climate models (CMs) included in the PCMDI/CMIP3 dataset in
pre-industrial climate conditions. It is found that about 90% of the material
entropy production is due to vertical processes such as convection, whereas the
large scale meridional heat transport contributes only about 10%. The total
material entropy production is typically 55 mWK-1m-2, with discrepancies of the
order of 5% and CMs' baroclinic efficiencies are clustered around 0.055. When
looking at the variability and co-variability of the considered thermodynamical
quantities, the agreement among CMs is worse, suggesting that the description
of feedbacks is more uncertain.
| [
{
"version": "v1",
"created": "Sun, 31 Jan 2010 20:51:03 GMT"
},
{
"version": "v2",
"created": "Mon, 15 Feb 2010 15:37:11 GMT"
},
{
"version": "v3",
"created": "Wed, 3 Nov 2010 13:45:23 GMT"
},
{
"version": "v4",
"created": "Tue, 7 Jun 2011 06:57:00 GMT"
}
] | 2015-05-18T00:00:00 | [
[
"Lucarini",
"Valerio",
""
],
[
"Fraedrich",
"Klaus",
""
],
[
"Ragone",
"Francesco",
""
]
] | TITLE: New Results on the Thermodynamical Properties of the Climate System
ABSTRACT: In this paper we exploit two equivalent formulations of the average rate of
material entropy production in a planetary system to propose an approximate
splitting between contributions due to vertical and eminently horizontal
processes. Our approach is based only upon 2D radiative fields at the surface
and at the top of atmosphere of a general planetary body. Using 2D fields at
the top of atmosphere alone, we derive lower bounds to the rate of material
entropy production and to the intensity of the Lorenz energy cycle. By
introducing a measure of the efficiency of the planetary system with respect to
horizontal thermodynamical processes, we provide insight on a previous
intuition on the possibility of defining a baroclinic heat engine extracting
work from the meridional heat flux. The approximate formula of the material
entropy production is verified and used for studying the global thermodynamic
properties of climate models (CMs) included in the PCMDI/CMIP3 dataset in
pre-industrial climate conditions. It is found that about 90% of the material
entropy production is due to vertical processes such as convection, whereas the
large scale meridional heat transport contributes only about 10%. The total
material entropy production is typically 55 mWK-1m-2, with discrepancies of the
order of 5% and CMs' baroclinic efficiencies are clustered around 0.055. When
looking at the variability and co-variability of the considered thermodynamical
quantities, the agreement among CMs is worse, suggesting that the description
of feedbacks is more uncertain.
| no_new_dataset | 0.948965 |
1003.0368 | Ana M Mancho | Carolina Mendoza, Ana M Mancho | The hidden geometry of ocean flows | 4 pages, 4 figures | Physical Review Letters 105 (3) 038501 (2010) | 10.1103/PhysRevLett.105.038501 | null | nlin.CD physics.ao-ph physics.flu-dyn | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We introduce a new global Lagrangian descriptor that is applied to flows with
general time dependence (altimetric datasets). It succeeds in detecting
simultaneously, with great accuracy, invariant manifolds, hyperbolic and
non-hyperbolic flow regions.
| [
{
"version": "v1",
"created": "Mon, 1 Mar 2010 15:05:17 GMT"
},
{
"version": "v2",
"created": "Thu, 15 Jul 2010 15:03:00 GMT"
}
] | 2015-05-18T00:00:00 | [
[
"Mendoza",
"Carolina",
""
],
[
"Mancho",
"Ana M",
""
]
] | TITLE: The hidden geometry of ocean flows
ABSTRACT: We introduce a new global Lagrangian descriptor that is applied to flows with
general time dependence (altimetric datasets). It succeeds in detecting
simultaneously, with great accuracy, invariant manifolds, hyperbolic and
non-hyperbolic flow regions.
| no_new_dataset | 0.952086 |
1003.0533 | Luca Sorriso-Valvo | V. Carbone, R. Marino, L. Sorriso-Valvo, A. Noullez, R. Bruno | Scaling Laws of Turbulence and Heating of Fast SolarWind: The Role of
Density Fluctuations | null | Rhys. Rev. Lett. 103, 061102, 2009 | 10.1103/PhysRevLett.103.061102 | null | physics.space-ph physics.flu-dyn physics.plasm-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Incompressible and isotropic magnetohydrodynamic turbulence in plasms can be
described by an exact relation for the energy flux through the scales. This
Yaglom-like scaling law has been recently observed in the solar wind above the
solar poles observed by the Ulysses spacecraft, where the turbulence is in an
Alfv\'enic state. An analogous phenomenological scaling law, suitably modified
to take into account compressible fluctuations, is observed more frequently in
the same dataset. Large scale density fluctuations, despite their low
amplitude, play thus a crucial role in the basic scaling properties of
turbulence. The turbulent cascade rate in the compressive case can moreover
supply the energy dissipation needed to account for the local heating of the
non-adiabatic solar wind.
| [
{
"version": "v1",
"created": "Tue, 2 Mar 2010 09:23:20 GMT"
}
] | 2015-05-18T00:00:00 | [
[
"Carbone",
"V.",
""
],
[
"Marino",
"R.",
""
],
[
"Sorriso-Valvo",
"L.",
""
],
[
"Noullez",
"A.",
""
],
[
"Bruno",
"R.",
""
]
] | TITLE: Scaling Laws of Turbulence and Heating of Fast SolarWind: The Role of
Density Fluctuations
ABSTRACT: Incompressible and isotropic magnetohydrodynamic turbulence in plasms can be
described by an exact relation for the energy flux through the scales. This
Yaglom-like scaling law has been recently observed in the solar wind above the
solar poles observed by the Ulysses spacecraft, where the turbulence is in an
Alfv\'enic state. An analogous phenomenological scaling law, suitably modified
to take into account compressible fluctuations, is observed more frequently in
the same dataset. Large scale density fluctuations, despite their low
amplitude, play thus a crucial role in the basic scaling properties of
turbulence. The turbulent cascade rate in the compressive case can moreover
supply the energy dissipation needed to account for the local heating of the
non-adiabatic solar wind.
| no_new_dataset | 0.949949 |
1003.1931 | Zi-Ke Zhang Mr. | Zi-Ke Zhang, Chuang Liu | Hypergraph model of social tagging networks | 7 pages,7 figures, 32 references | J. Stat. Mech. (2010) P100005 | 10.1088/1742-5468/2010/10/P10005 | null | physics.soc-ph cs.IR | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The past few years have witnessed the great success of a new family of
paradigms, so-called folksonomy, which allows users to freely associate tags to
resources and efficiently manage them. In order to uncover the underlying
structures and user behaviors in folksonomy, in this paper, we propose an
evolutionary hypergrah model to explain the emerging statistical properties.
The present model introduces a novel mechanism that one can not only assign
tags to resources, but also retrieve resources via collaborative tags. We then
compare the model with a real-world dataset: \emph{Del.icio.us}. Indeed, the
present model shows considerable agreement with the empirical data in following
aspects: power-law hyperdegree distributions, negtive correlation between
clustering coefficients and hyperdegrees, and small average distances.
Furthermore, the model indicates that most tagging behaviors are motivated by
labeling tags to resources, and tags play a significant role in effectively
retrieving interesting resources and making acquaintance with congenial
friends. The proposed model may shed some light on the in-depth understanding
of the structure and function of folksonomy.
| [
{
"version": "v1",
"created": "Tue, 9 Mar 2010 17:03:41 GMT"
}
] | 2015-05-18T00:00:00 | [
[
"Zhang",
"Zi-Ke",
""
],
[
"Liu",
"Chuang",
""
]
] | TITLE: Hypergraph model of social tagging networks
ABSTRACT: The past few years have witnessed the great success of a new family of
paradigms, so-called folksonomy, which allows users to freely associate tags to
resources and efficiently manage them. In order to uncover the underlying
structures and user behaviors in folksonomy, in this paper, we propose an
evolutionary hypergrah model to explain the emerging statistical properties.
The present model introduces a novel mechanism that one can not only assign
tags to resources, but also retrieve resources via collaborative tags. We then
compare the model with a real-world dataset: \emph{Del.icio.us}. Indeed, the
present model shows considerable agreement with the empirical data in following
aspects: power-law hyperdegree distributions, negtive correlation between
clustering coefficients and hyperdegrees, and small average distances.
Furthermore, the model indicates that most tagging behaviors are motivated by
labeling tags to resources, and tags play a significant role in effectively
retrieving interesting resources and making acquaintance with congenial
friends. The proposed model may shed some light on the in-depth understanding
of the structure and function of folksonomy.
| no_new_dataset | 0.928797 |
1306.1461 | Bob Sturm | Bob L. Sturm | The GTZAN dataset: Its contents, its faults, their effects on
evaluation, and its future use | 29 pages, 7 figures, 6 tables, 128 references | null | 10.1080/09298215.2014.894533 | null | cs.SD | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The GTZAN dataset appears in at least 100 published works, and is the
most-used public dataset for evaluation in machine listening research for music
genre recognition (MGR). Our recent work, however, shows GTZAN has several
faults (repetitions, mislabelings, and distortions), which challenge the
interpretability of any result derived using it. In this article, we disprove
the claims that all MGR systems are affected in the same ways by these faults,
and that the performances of MGR systems in GTZAN are still meaningfully
comparable since they all face the same faults. We identify and analyze the
contents of GTZAN, and provide a catalog of its faults. We review how GTZAN has
been used in MGR research, and find few indications that its faults have been
known and considered. Finally, we rigorously study the effects of its faults on
evaluating five different MGR systems. The lesson is not to banish GTZAN, but
to use it with consideration of its contents.
| [
{
"version": "v1",
"created": "Thu, 6 Jun 2013 16:30:44 GMT"
},
{
"version": "v2",
"created": "Fri, 7 Jun 2013 16:57:39 GMT"
}
] | 2015-05-18T00:00:00 | [
[
"Sturm",
"Bob L.",
""
]
] | TITLE: The GTZAN dataset: Its contents, its faults, their effects on
evaluation, and its future use
ABSTRACT: The GTZAN dataset appears in at least 100 published works, and is the
most-used public dataset for evaluation in machine listening research for music
genre recognition (MGR). Our recent work, however, shows GTZAN has several
faults (repetitions, mislabelings, and distortions), which challenge the
interpretability of any result derived using it. In this article, we disprove
the claims that all MGR systems are affected in the same ways by these faults,
and that the performances of MGR systems in GTZAN are still meaningfully
comparable since they all face the same faults. We identify and analyze the
contents of GTZAN, and provide a catalog of its faults. We review how GTZAN has
been used in MGR research, and find few indications that its faults have been
known and considered. Finally, we rigorously study the effects of its faults on
evaluating five different MGR systems. The lesson is not to banish GTZAN, but
to use it with consideration of its contents.
| no_new_dataset | 0.919931 |
1505.03358 | Miriam Redi | Rossano Schifanella, Miriam Redi, Luca Aiello | An Image is Worth More than a Thousand Favorites: Surfacing the Hidden
Beauty of Flickr Pictures | ICWSM 2015 | null | null | null | cs.SI cs.CV cs.CY cs.MM | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The dynamics of attention in social media tend to obey power laws. Attention
concentrates on a relatively small number of popular items and neglecting the
vast majority of content produced by the crowd. Although popularity can be an
indication of the perceived value of an item within its community, previous
research has hinted to the fact that popularity is distinct from intrinsic
quality. As a result, content with low visibility but high quality lurks in the
tail of the popularity distribution. This phenomenon can be particularly
evident in the case of photo-sharing communities, where valuable photographers
who are not highly engaged in online social interactions contribute with
high-quality pictures that remain unseen. We propose to use a computer vision
method to surface beautiful pictures from the immense pool of
near-zero-popularity items, and we test it on a large dataset of
creative-commons photos on Flickr. By gathering a large crowdsourced ground
truth of aesthetics scores for Flickr images, we show that our method retrieves
photos whose median perceived beauty score is equal to the most popular ones,
and whose average is lower by only 1.5%.
| [
{
"version": "v1",
"created": "Wed, 13 May 2015 12:40:24 GMT"
},
{
"version": "v2",
"created": "Fri, 15 May 2015 10:05:34 GMT"
}
] | 2015-05-18T00:00:00 | [
[
"Schifanella",
"Rossano",
""
],
[
"Redi",
"Miriam",
""
],
[
"Aiello",
"Luca",
""
]
] | TITLE: An Image is Worth More than a Thousand Favorites: Surfacing the Hidden
Beauty of Flickr Pictures
ABSTRACT: The dynamics of attention in social media tend to obey power laws. Attention
concentrates on a relatively small number of popular items and neglecting the
vast majority of content produced by the crowd. Although popularity can be an
indication of the perceived value of an item within its community, previous
research has hinted to the fact that popularity is distinct from intrinsic
quality. As a result, content with low visibility but high quality lurks in the
tail of the popularity distribution. This phenomenon can be particularly
evident in the case of photo-sharing communities, where valuable photographers
who are not highly engaged in online social interactions contribute with
high-quality pictures that remain unseen. We propose to use a computer vision
method to surface beautiful pictures from the immense pool of
near-zero-popularity items, and we test it on a large dataset of
creative-commons photos on Flickr. By gathering a large crowdsourced ground
truth of aesthetics scores for Flickr images, we show that our method retrieves
photos whose median perceived beauty score is equal to the most popular ones,
and whose average is lower by only 1.5%.
| no_new_dataset | 0.93233 |
1505.03873 | Kevin Tang | Kevin Tang and Manohar Paluri and Li Fei-Fei and Rob Fergus and
Lubomir Bourdev | Improving Image Classification with Location Context | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | With the widespread availability of cellphones and cameras that have GPS
capabilities, it is common for images being uploaded to the Internet today to
have GPS coordinates associated with them. In addition to research that tries
to predict GPS coordinates from visual features, this also opens up the door to
problems that are conditioned on the availability of GPS coordinates. In this
work, we tackle the problem of performing image classification with location
context, in which we are given the GPS coordinates for images in both the train
and test phases. We explore different ways of encoding and extracting features
from the GPS coordinates, and show how to naturally incorporate these features
into a Convolutional Neural Network (CNN), the current state-of-the-art for
most image classification and recognition problems. We also show how it is
possible to simultaneously learn the optimal pooling radii for a subset of our
features within the CNN framework. To evaluate our model and to help promote
research in this area, we identify a set of location-sensitive concepts and
annotate a subset of the Yahoo Flickr Creative Commons 100M dataset that has
GPS coordinates with these concepts, which we make publicly available. By
leveraging location context, we are able to achieve almost a 7% gain in mean
average precision.
| [
{
"version": "v1",
"created": "Thu, 14 May 2015 20:13:34 GMT"
}
] | 2015-05-18T00:00:00 | [
[
"Tang",
"Kevin",
""
],
[
"Paluri",
"Manohar",
""
],
[
"Fei-Fei",
"Li",
""
],
[
"Fergus",
"Rob",
""
],
[
"Bourdev",
"Lubomir",
""
]
] | TITLE: Improving Image Classification with Location Context
ABSTRACT: With the widespread availability of cellphones and cameras that have GPS
capabilities, it is common for images being uploaded to the Internet today to
have GPS coordinates associated with them. In addition to research that tries
to predict GPS coordinates from visual features, this also opens up the door to
problems that are conditioned on the availability of GPS coordinates. In this
work, we tackle the problem of performing image classification with location
context, in which we are given the GPS coordinates for images in both the train
and test phases. We explore different ways of encoding and extracting features
from the GPS coordinates, and show how to naturally incorporate these features
into a Convolutional Neural Network (CNN), the current state-of-the-art for
most image classification and recognition problems. We also show how it is
possible to simultaneously learn the optimal pooling radii for a subset of our
features within the CNN framework. To evaluate our model and to help promote
research in this area, we identify a set of location-sensitive concepts and
annotate a subset of the Yahoo Flickr Creative Commons 100M dataset that has
GPS coordinates with these concepts, which we make publicly available. By
leveraging location context, we are able to achieve almost a 7% gain in mean
average precision.
| no_new_dataset | 0.946794 |
1503.03957 | Prabhjot Singh | Prabhjot Singh, Sumit Dhawan, Shubham Agarwal, Narina Thakur | Implementation of an efficient Fuzzy Logic based Information Retrieval
System | arXiv admin note: substantial text overlap with
http://ntz-develop.blogspot.in/ ,
http://www.micsymposium.org/mics2012/submissions/mics2012_submission_8.pdf ,
http://www.slideshare.net/JeffreyStricklandPhD/predictive-modeling-and-analytics-selectchapters-41304405
by other authors | null | null | null | cs.IR | http://creativecommons.org/licenses/by/3.0/ | This paper exemplifies the implementation of an efficient Information
Retrieval (IR) System to compute the similarity between a dataset and a query
using Fuzzy Logic. TREC dataset has been used for the same purpose. The dataset
is parsed to generate keywords index which is used for the similarity
comparison with the user query. Each query is assigned a score value based on
its fuzzy similarity with the index keywords. The relevant documents are
retrieved based on the score value. The performance and accuracy of the
proposed fuzzy similarity model is compared with Cosine similarity model using
Precision-Recall curves. The results prove the dominance of Fuzzy Similarity
based IR system.
| [
{
"version": "v1",
"created": "Fri, 13 Mar 2015 05:21:02 GMT"
}
] | 2015-05-15T00:00:00 | [
[
"Singh",
"Prabhjot",
""
],
[
"Dhawan",
"Sumit",
""
],
[
"Agarwal",
"Shubham",
""
],
[
"Thakur",
"Narina",
""
]
] | TITLE: Implementation of an efficient Fuzzy Logic based Information Retrieval
System
ABSTRACT: This paper exemplifies the implementation of an efficient Information
Retrieval (IR) System to compute the similarity between a dataset and a query
using Fuzzy Logic. TREC dataset has been used for the same purpose. The dataset
is parsed to generate keywords index which is used for the similarity
comparison with the user query. Each query is assigned a score value based on
its fuzzy similarity with the index keywords. The relevant documents are
retrieved based on the score value. The performance and accuracy of the
proposed fuzzy similarity model is compared with Cosine similarity model using
Precision-Recall curves. The results prove the dominance of Fuzzy Similarity
based IR system.
| no_new_dataset | 0.949482 |
1504.05457 | Yury Kashnitsky | Yury Kashnitsky, Sergei O. Kuznetsov | Graphlet-based lazy associative graph classification | This paper has been withdrawn by the author due to the incomplete set
of necessary definitions and experiments | null | null | null | cs.DS cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The paper addresses the graph classification problem and introduces a
modification of the lazy associative classification method to efficiently
handle intersections of graphs. Graph intersections are approximated with all
common subgraphs up to a fixed size similarly to what is done with graphlet
kernels. We explain the idea of the algorithm with a toy example and describe
our experiments with a predictive toxicology dataset.
| [
{
"version": "v1",
"created": "Tue, 21 Apr 2015 15:12:45 GMT"
},
{
"version": "v2",
"created": "Wed, 13 May 2015 20:46:47 GMT"
}
] | 2015-05-15T00:00:00 | [
[
"Kashnitsky",
"Yury",
""
],
[
"Kuznetsov",
"Sergei O.",
""
]
] | TITLE: Graphlet-based lazy associative graph classification
ABSTRACT: The paper addresses the graph classification problem and introduces a
modification of the lazy associative classification method to efficiently
handle intersections of graphs. Graph intersections are approximated with all
common subgraphs up to a fixed size similarly to what is done with graphlet
kernels. We explain the idea of the algorithm with a toy example and describe
our experiments with a predictive toxicology dataset.
| no_new_dataset | 0.940953 |
1505.03560 | Sebastiano Stramaglia | Ibai Diez, Asier Erramuzpe, Inaki Escudero, Beatriz Mateos, Alberto
Cabrera, Daniele Marinazzo, Ernesto J. Sanz-Arigita, Sebastiano Stramaglia
and Jesus M. Cortes | Information flow between resting state networks | 47 pages, 5 figures, 4 tables, 3 supplementary figures. Accepted for
publication in Brain Connectivity in its current form | null | null | null | q-bio.NC physics.data-an q-bio.QM | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The resting brain dynamics self-organizes into a finite number of correlated
patterns known as resting state networks (RSNs). It is well known that
techniques like independent component analysis can separate the brain activity
at rest to provide such RSNs, but the specific pattern of interaction between
RSNs is not yet fully understood. To this aim, we propose here a novel method
to compute the information flow (IF) between different RSNs from resting state
magnetic resonance imaging. After haemodynamic response function blind
deconvolution of all voxel signals, and under the hypothesis that RSNs define
regions of interest, our method first uses principal component analysis to
reduce dimensionality in each RSN to next compute IF (estimated here in terms
of Transfer Entropy) between the different RSNs by systematically increasing k
(the number of principal components used in the calculation). When k = 1, this
method is equivalent to computing IF using the average of all voxel activities
in each RSN. For k greater than one our method calculates the k-multivariate IF
between the different RSNs. We find that the average IF among RSNs is
dimension-dependent, increasing from k =1 (i.e., the average voxels activity)
up to a maximum occurring at k =5 to finally decay to zero for k greater than
10. This suggests that a small number of components (close to 5) is sufficient
to describe the IF pattern between RSNs. Our method - addressing differences in
IF between RSNs for any generic data - can be used for group comparison in
health or disease. To illustrate this, we have calculated the interRSNs IF in a
dataset of Alzheimer's Disease (AD) to find that the most significant
differences between AD and controls occurred for k =2, in addition to AD
showing increased IF w.r.t. controls.
| [
{
"version": "v1",
"created": "Wed, 13 May 2015 21:45:44 GMT"
}
] | 2015-05-15T00:00:00 | [
[
"Diez",
"Ibai",
""
],
[
"Erramuzpe",
"Asier",
""
],
[
"Escudero",
"Inaki",
""
],
[
"Mateos",
"Beatriz",
""
],
[
"Cabrera",
"Alberto",
""
],
[
"Marinazzo",
"Daniele",
""
],
[
"Sanz-Arigita",
"Ernesto J.",
""
],
[
"Stramaglia",
"Sebastiano",
""
],
[
"Cortes",
"Jesus M.",
""
]
] | TITLE: Information flow between resting state networks
ABSTRACT: The resting brain dynamics self-organizes into a finite number of correlated
patterns known as resting state networks (RSNs). It is well known that
techniques like independent component analysis can separate the brain activity
at rest to provide such RSNs, but the specific pattern of interaction between
RSNs is not yet fully understood. To this aim, we propose here a novel method
to compute the information flow (IF) between different RSNs from resting state
magnetic resonance imaging. After haemodynamic response function blind
deconvolution of all voxel signals, and under the hypothesis that RSNs define
regions of interest, our method first uses principal component analysis to
reduce dimensionality in each RSN to next compute IF (estimated here in terms
of Transfer Entropy) between the different RSNs by systematically increasing k
(the number of principal components used in the calculation). When k = 1, this
method is equivalent to computing IF using the average of all voxel activities
in each RSN. For k greater than one our method calculates the k-multivariate IF
between the different RSNs. We find that the average IF among RSNs is
dimension-dependent, increasing from k =1 (i.e., the average voxels activity)
up to a maximum occurring at k =5 to finally decay to zero for k greater than
10. This suggests that a small number of components (close to 5) is sufficient
to describe the IF pattern between RSNs. Our method - addressing differences in
IF between RSNs for any generic data - can be used for group comparison in
health or disease. To illustrate this, we have calculated the interRSNs IF in a
dataset of Alzheimer's Disease (AD) to find that the most significant
differences between AD and controls occurred for k =2, in addition to AD
showing increased IF w.r.t. controls.
| no_new_dataset | 0.947381 |
1505.03581 | Ali Borji | Ali Borji, Laurent Itti | CAT2000: A Large Scale Fixation Dataset for Boosting Saliency Research | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Saliency modeling has been an active research area in computer vision for
about two decades. Existing state of the art models perform very well in
predicting where people look in natural scenes. There is, however, the risk
that these models may have been overfitting themselves to available small scale
biased datasets, thus trapping the progress in a local minimum. To gain a
deeper insight regarding current issues in saliency modeling and to better
gauge progress, we recorded eye movements of 120 observers while they freely
viewed a large number of naturalistic and artificial images. Our stimuli
includes 4000 images; 200 from each of 20 categories covering different types
of scenes such as Cartoons, Art, Objects, Low resolution images, Indoor,
Outdoor, Jumbled, Random, and Line drawings. We analyze some basic properties
of this dataset and compare some successful models. We believe that our dataset
opens new challenges for the next generation of saliency models and helps
conduct behavioral studies on bottom-up visual attention.
| [
{
"version": "v1",
"created": "Thu, 14 May 2015 00:34:43 GMT"
}
] | 2015-05-15T00:00:00 | [
[
"Borji",
"Ali",
""
],
[
"Itti",
"Laurent",
""
]
] | TITLE: CAT2000: A Large Scale Fixation Dataset for Boosting Saliency Research
ABSTRACT: Saliency modeling has been an active research area in computer vision for
about two decades. Existing state of the art models perform very well in
predicting where people look in natural scenes. There is, however, the risk
that these models may have been overfitting themselves to available small scale
biased datasets, thus trapping the progress in a local minimum. To gain a
deeper insight regarding current issues in saliency modeling and to better
gauge progress, we recorded eye movements of 120 observers while they freely
viewed a large number of naturalistic and artificial images. Our stimuli
includes 4000 images; 200 from each of 20 categories covering different types
of scenes such as Cartoons, Art, Objects, Low resolution images, Indoor,
Outdoor, Jumbled, Random, and Line drawings. We analyze some basic properties
of this dataset and compare some successful models. We believe that our dataset
opens new challenges for the next generation of saliency models and helps
conduct behavioral studies on bottom-up visual attention.
| new_dataset | 0.961929 |
0909.3893 | Cesar Hidalgo | Cesar A. Hidalgo, Nicholas Blumm, Albert-Laszlo Barabasi, Nicholas
Christakis | A dynamic network approach for the study of human phenotypes | 28 pages (double space), 6 figures | null | 10.1371/journal.pcbi.1000353 | null | physics.bio-ph physics.data-an physics.med-ph q-bio.QM | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The use of networks to integrate different genetic, proteomic, and metabolic
datasets has been proposed as a viable path toward elucidating the origins of
specific diseases. Here we introduce a new phenotypic database summarizing
correlations obtained from the disease history of more than 30 million patients
in a Phenotypic Disease Network (PDN). We present evidence that the structure
of the PDN is relevant to the understanding of illness progression by showing
that (1) patients develop diseases close in the network to those they already
have; (2) the progression of disease along the links of the network is
different for patients of different genders and ethnicities; (3) patients
diagnosed with diseases which are more highly connected in the PDN tend to die
sooner than those affected by less connected diseases; and (4) diseases that
tend to be preceded by others in the PDN tend to be more connected than
diseases that precede other illnesses, and are associated with higher degrees
of mortality. Our findings show that disease progression can be represented and
studied using network methods, offering the potential to enhance our
understanding of the origin and evolution of human diseases. The dataset
introduced here, released concurrently with this publication, represents the
largest relational phenotypic resource publicly available to the research
community.
| [
{
"version": "v1",
"created": "Tue, 22 Sep 2009 02:25:42 GMT"
}
] | 2015-05-14T00:00:00 | [
[
"Hidalgo",
"Cesar A.",
""
],
[
"Blumm",
"Nicholas",
""
],
[
"Barabasi",
"Albert-Laszlo",
""
],
[
"Christakis",
"Nicholas",
""
]
] | TITLE: A dynamic network approach for the study of human phenotypes
ABSTRACT: The use of networks to integrate different genetic, proteomic, and metabolic
datasets has been proposed as a viable path toward elucidating the origins of
specific diseases. Here we introduce a new phenotypic database summarizing
correlations obtained from the disease history of more than 30 million patients
in a Phenotypic Disease Network (PDN). We present evidence that the structure
of the PDN is relevant to the understanding of illness progression by showing
that (1) patients develop diseases close in the network to those they already
have; (2) the progression of disease along the links of the network is
different for patients of different genders and ethnicities; (3) patients
diagnosed with diseases which are more highly connected in the PDN tend to die
sooner than those affected by less connected diseases; and (4) diseases that
tend to be preceded by others in the PDN tend to be more connected than
diseases that precede other illnesses, and are associated with higher degrees
of mortality. Our findings show that disease progression can be represented and
studied using network methods, offering the potential to enhance our
understanding of the origin and evolution of human diseases. The dataset
introduced here, released concurrently with this publication, represents the
largest relational phenotypic resource publicly available to the research
community.
| new_dataset | 0.960398 |
0910.0767 | Serena Bradde | M. Bailly-Bechet, S. Bradde, A. Braunstein, A. Flaxman, L. Foini, R.
Zecchina | Clustering with shallow trees | 11 pages, 7 figures | J. Stat. Mech. (2009) P12010 | 10.1088/1742-5468/2009/12/P12010 | null | cond-mat.dis-nn cs.DS q-bio.QM | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We propose a new method for hierarchical clustering based on the optimisation
of a cost function over trees of limited depth, and we derive a
message--passing method that allows to solve it efficiently. The method and
algorithm can be interpreted as a natural interpolation between two well-known
approaches, namely single linkage and the recently presented Affinity
Propagation. We analyze with this general scheme three biological/medical
structured datasets (human population based on genetic information, proteins
based on sequences and verbal autopsies) and show that the interpolation
technique provides new insight.
| [
{
"version": "v1",
"created": "Mon, 5 Oct 2009 14:13:25 GMT"
},
{
"version": "v2",
"created": "Mon, 23 Nov 2009 15:44:29 GMT"
}
] | 2015-05-14T00:00:00 | [
[
"Bailly-Bechet",
"M.",
""
],
[
"Bradde",
"S.",
""
],
[
"Braunstein",
"A.",
""
],
[
"Flaxman",
"A.",
""
],
[
"Foini",
"L.",
""
],
[
"Zecchina",
"R.",
""
]
] | TITLE: Clustering with shallow trees
ABSTRACT: We propose a new method for hierarchical clustering based on the optimisation
of a cost function over trees of limited depth, and we derive a
message--passing method that allows to solve it efficiently. The method and
algorithm can be interpreted as a natural interpolation between two well-known
approaches, namely single linkage and the recently presented Affinity
Propagation. We analyze with this general scheme three biological/medical
structured datasets (human population based on genetic information, proteins
based on sequences and verbal autopsies) and show that the interpolation
technique provides new insight.
| no_new_dataset | 0.948537 |
0912.2826 | Floriana Gargiulo | Floriana Gargiulo, Sonia Ternes, Sylvie Huet, Guillaume Deffuant | An iterative approach for generating statistically realistic populations
of households | 16 oages, 11 figures | null | 10.1371/journal.pone.0008828 | null | cs.MA cs.CY | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Background: Many different simulation frameworks, in different topics, need
to treat realistic datasets to initialize and calibrate the system. A precise
reproduction of initial states is extremely important to obtain reliable
forecast from the model. Methodology/Principal Findings: This paper proposes an
algorithm to create an artificial population where individuals are described by
their age, and are gathered in households respecting a variety of statistical
constraints (distribution of household types, sizes, age of household head,
difference of age between partners and among parents and children). Such a
population is often the initial state of microsimulation or (agent)
individual-based models. To get a realistic distribution of households is often
very important, because this distribution has an impact on the demographic
evolution. Usual techniques from microsimulation approach cross different
sources of aggregated data for generating individuals. In our case the number
of combinations of different households (types, sizes, age of participants)
makes it computationally difficult to use directly such methods. Hence we
developed a specific algorithm to make the problem more easily tractable.
Conclusions/Significance: We generate the populations of two pilot
municipalities in Auvergne region (France), to illustrate the approach. The
generated populations show a good agreement with the available statistical
datasets (not used for the generation) and are obtained in a reasonable
computational time.
| [
{
"version": "v1",
"created": "Tue, 15 Dec 2009 09:28:33 GMT"
}
] | 2015-05-14T00:00:00 | [
[
"Gargiulo",
"Floriana",
""
],
[
"Ternes",
"Sonia",
""
],
[
"Huet",
"Sylvie",
""
],
[
"Deffuant",
"Guillaume",
""
]
] | TITLE: An iterative approach for generating statistically realistic populations
of households
ABSTRACT: Background: Many different simulation frameworks, in different topics, need
to treat realistic datasets to initialize and calibrate the system. A precise
reproduction of initial states is extremely important to obtain reliable
forecast from the model. Methodology/Principal Findings: This paper proposes an
algorithm to create an artificial population where individuals are described by
their age, and are gathered in households respecting a variety of statistical
constraints (distribution of household types, sizes, age of household head,
difference of age between partners and among parents and children). Such a
population is often the initial state of microsimulation or (agent)
individual-based models. To get a realistic distribution of households is often
very important, because this distribution has an impact on the demographic
evolution. Usual techniques from microsimulation approach cross different
sources of aggregated data for generating individuals. In our case the number
of combinations of different households (types, sizes, age of participants)
makes it computationally difficult to use directly such methods. Hence we
developed a specific algorithm to make the problem more easily tractable.
Conclusions/Significance: We generate the populations of two pilot
municipalities in Auvergne region (France), to illustrate the approach. The
generated populations show a good agreement with the available statistical
datasets (not used for the generation) and are obtained in a reasonable
computational time.
| no_new_dataset | 0.949529 |
1503.05214 | Tarek Elgamal | Tarek Elgamal, Mohamed Hefeeda | Analysis of PCA Algorithms in Distributed Environments | null | null | null | null | cs.DC cs.LG cs.NA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Classical machine learning algorithms often face scalability bottlenecks when
they are applied to large-scale data. Such algorithms were designed to work
with small data that is assumed to fit in the memory of one machine. In this
report, we analyze different methods for computing an important machine learing
algorithm, namely Principal Component Analysis (PCA), and we comment on its
limitations in supporting large datasets. The methods are analyzed and compared
across two important metrics: time complexity and communication complexity. We
consider the worst-case scenarios for both metrics, and we identify the
software libraries that implement each method. The analysis in this report
helps researchers and engineers in (i) understanding the main bottlenecks for
scalability in different PCA algorithms, (ii) choosing the most appropriate
method and software library for a given application and data set
characteristics, and (iii) designing new scalable PCA algorithms.
| [
{
"version": "v1",
"created": "Tue, 17 Mar 2015 20:38:15 GMT"
},
{
"version": "v2",
"created": "Wed, 13 May 2015 12:05:02 GMT"
}
] | 2015-05-14T00:00:00 | [
[
"Elgamal",
"Tarek",
""
],
[
"Hefeeda",
"Mohamed",
""
]
] | TITLE: Analysis of PCA Algorithms in Distributed Environments
ABSTRACT: Classical machine learning algorithms often face scalability bottlenecks when
they are applied to large-scale data. Such algorithms were designed to work
with small data that is assumed to fit in the memory of one machine. In this
report, we analyze different methods for computing an important machine learing
algorithm, namely Principal Component Analysis (PCA), and we comment on its
limitations in supporting large datasets. The methods are analyzed and compared
across two important metrics: time complexity and communication complexity. We
consider the worst-case scenarios for both metrics, and we identify the
software libraries that implement each method. The analysis in this report
helps researchers and engineers in (i) understanding the main bottlenecks for
scalability in different PCA algorithms, (ii) choosing the most appropriate
method and software library for a given application and data set
characteristics, and (iii) designing new scalable PCA algorithms.
| no_new_dataset | 0.952926 |
1505.01919 | Puja Gupta | Puja Gupta | Characterization of Performance Anomalies in Hadoop | null | null | null | null | cs.DC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | With the huge variety of data and equally large-scale systems, there is not a
unique execution setting for these systems which can guarantee the best
performance for each query. In this project, we tried so study the impact of
different execution settings on execution time of workloads by varying them one
at a time. Using the data from these experiments, a decision tree was built
where each internal node represents the execution parameter, each branch
represents value chosen for the parameter and each leaf node represents a range
for execution time in minutes. The attribute in the decision tree to split the
dataset on is selected based on the maximum information gain or lowest entropy.
Once the tree is trained with the training samples, this tree can be used to
get approximate range for the expected execution time. When the actual
execution time differs from this expected value, a performance anomaly can be
detected. For a test dataset with 400 samples, 99% of samples had actual
execution time in the range predicted time by the decision tree. Also on
analyzing the constructed tree, an idea about what configuration can give
better performance for a given workload can be obtained. Initial experiments
suggest that the impact an execution parameter can have on the target attribute
(here execution time) can be related to the distance of that feature node from
the root of the constructed decision tree. From initial results the percent
change in the values of the target attribute for various value of the feature
node which is closer to the root is 6 times larger than when that same iii
feature node is away from the root node. This observation will depend on how
well the decision tree was trained and may not be true for every case.
| [
{
"version": "v1",
"created": "Fri, 8 May 2015 03:24:33 GMT"
},
{
"version": "v2",
"created": "Wed, 13 May 2015 03:50:26 GMT"
}
] | 2015-05-14T00:00:00 | [
[
"Gupta",
"Puja",
""
]
] | TITLE: Characterization of Performance Anomalies in Hadoop
ABSTRACT: With the huge variety of data and equally large-scale systems, there is not a
unique execution setting for these systems which can guarantee the best
performance for each query. In this project, we tried so study the impact of
different execution settings on execution time of workloads by varying them one
at a time. Using the data from these experiments, a decision tree was built
where each internal node represents the execution parameter, each branch
represents value chosen for the parameter and each leaf node represents a range
for execution time in minutes. The attribute in the decision tree to split the
dataset on is selected based on the maximum information gain or lowest entropy.
Once the tree is trained with the training samples, this tree can be used to
get approximate range for the expected execution time. When the actual
execution time differs from this expected value, a performance anomaly can be
detected. For a test dataset with 400 samples, 99% of samples had actual
execution time in the range predicted time by the decision tree. Also on
analyzing the constructed tree, an idea about what configuration can give
better performance for a given workload can be obtained. Initial experiments
suggest that the impact an execution parameter can have on the target attribute
(here execution time) can be related to the distance of that feature node from
the root of the constructed decision tree. From initial results the percent
change in the values of the target attribute for various value of the feature
node which is closer to the root is 6 times larger than when that same iii
feature node is away from the root node. This observation will depend on how
well the decision tree was trained and may not be true for every case.
| no_new_dataset | 0.955361 |
1505.02212 | Yakir Reshef | Yakir A. Reshef, David N. Reshef, Pardis C. Sabeti, Michael M.
Mitzenmacher | Equitability, interval estimation, and statistical power | Yakir A. Reshef and David N. Reshef are co-first authors, Pardis C.
Sabeti and Michael M. Mitzenmacher are co-last authors. This paper, together
with arXiv:1505.02212, subsumes arXiv:1408.4908 | null | null | null | math.ST cs.LG q-bio.QM stat.ME stat.ML stat.TH | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | For analysis of a high-dimensional dataset, a common approach is to test a
null hypothesis of statistical independence on all variable pairs using a
non-parametric measure of dependence. However, because this approach attempts
to identify any non-trivial relationship no matter how weak, it often
identifies too many relationships to be useful. What is needed is a way of
identifying a smaller set of relationships that merit detailed further
analysis.
Here we formally present and characterize equitability, a property of
measures of dependence that aims to overcome this challenge. Notionally, an
equitable statistic is a statistic that, given some measure of noise, assigns
similar scores to equally noisy relationships of different types [Reshef et al.
2011]. We begin by formalizing this idea via a new object called the
interpretable interval, which functions as an interval estimate of the amount
of noise in a relationship of unknown type. We define an equitable statistic as
one with small interpretable intervals.
We then draw on the equivalence of interval estimation and hypothesis testing
to show that under moderate assumptions an equitable statistic is one that
yields well powered tests for distinguishing not only between trivial and
non-trivial relationships of all kinds but also between non-trivial
relationships of different strengths. This means that equitability allows us to
specify a threshold relationship strength $x_0$ and to search for relationships
of all kinds with strength greater than $x_0$. Thus, equitability can be
thought of as a strengthening of power against independence that enables
fruitful analysis of data sets with a small number of strong, interesting
relationships and a large number of weaker ones. We conclude with a
demonstration of how our two equivalent characterizations of equitability can
be used to evaluate the equitability of a statistic in practice.
| [
{
"version": "v1",
"created": "Sat, 9 May 2015 00:31:23 GMT"
},
{
"version": "v2",
"created": "Tue, 12 May 2015 20:05:17 GMT"
}
] | 2015-05-14T00:00:00 | [
[
"Reshef",
"Yakir A.",
""
],
[
"Reshef",
"David N.",
""
],
[
"Sabeti",
"Pardis C.",
""
],
[
"Mitzenmacher",
"Michael M.",
""
]
] | TITLE: Equitability, interval estimation, and statistical power
ABSTRACT: For analysis of a high-dimensional dataset, a common approach is to test a
null hypothesis of statistical independence on all variable pairs using a
non-parametric measure of dependence. However, because this approach attempts
to identify any non-trivial relationship no matter how weak, it often
identifies too many relationships to be useful. What is needed is a way of
identifying a smaller set of relationships that merit detailed further
analysis.
Here we formally present and characterize equitability, a property of
measures of dependence that aims to overcome this challenge. Notionally, an
equitable statistic is a statistic that, given some measure of noise, assigns
similar scores to equally noisy relationships of different types [Reshef et al.
2011]. We begin by formalizing this idea via a new object called the
interpretable interval, which functions as an interval estimate of the amount
of noise in a relationship of unknown type. We define an equitable statistic as
one with small interpretable intervals.
We then draw on the equivalence of interval estimation and hypothesis testing
to show that under moderate assumptions an equitable statistic is one that
yields well powered tests for distinguishing not only between trivial and
non-trivial relationships of all kinds but also between non-trivial
relationships of different strengths. This means that equitability allows us to
specify a threshold relationship strength $x_0$ and to search for relationships
of all kinds with strength greater than $x_0$. Thus, equitability can be
thought of as a strengthening of power against independence that enables
fruitful analysis of data sets with a small number of strong, interesting
relationships and a large number of weaker ones. We conclude with a
demonstration of how our two equivalent characterizations of equitability can
be used to evaluate the equitability of a statistic in practice.
| no_new_dataset | 0.938407 |
1505.02867 | Charles Mathy | Charles Mathy, Nate Derbinsky, Jos\'e Bento, Jonathan Rosenthal and
Jonathan Yedidia | The Boundary Forest Algorithm for Online Supervised and Unsupervised
Learning | 7 pages, 4 figs, 1 page supp. info | Proc. of the 29th AAAI Conference on Artificial Intelligence
(AAAI), 2864-2870. Austin, TX, USA. (2015) | null | null | cs.LG cs.DS cs.IR stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We describe a new instance-based learning algorithm called the Boundary
Forest (BF) algorithm, that can be used for supervised and unsupervised
learning. The algorithm builds a forest of trees whose nodes store previously
seen examples. It can be shown data points one at a time and updates itself
incrementally, hence it is naturally online. Few instance-based algorithms have
this property while being simultaneously fast, which the BF is. This is crucial
for applications where one needs to respond to input data in real time. The
number of children of each node is not set beforehand but obtained from the
training procedure, which makes the algorithm very flexible with regards to
what data manifolds it can learn. We test its generalization performance and
speed on a range of benchmark datasets and detail in which settings it
outperforms the state of the art. Empirically we find that training time scales
as O(DNlog(N)) and testing as O(Dlog(N)), where D is the dimensionality and N
the amount of data,
| [
{
"version": "v1",
"created": "Tue, 12 May 2015 03:45:11 GMT"
}
] | 2015-05-14T00:00:00 | [
[
"Mathy",
"Charles",
""
],
[
"Derbinsky",
"Nate",
""
],
[
"Bento",
"José",
""
],
[
"Rosenthal",
"Jonathan",
""
],
[
"Yedidia",
"Jonathan",
""
]
] | TITLE: The Boundary Forest Algorithm for Online Supervised and Unsupervised
Learning
ABSTRACT: We describe a new instance-based learning algorithm called the Boundary
Forest (BF) algorithm, that can be used for supervised and unsupervised
learning. The algorithm builds a forest of trees whose nodes store previously
seen examples. It can be shown data points one at a time and updates itself
incrementally, hence it is naturally online. Few instance-based algorithms have
this property while being simultaneously fast, which the BF is. This is crucial
for applications where one needs to respond to input data in real time. The
number of children of each node is not set beforehand but obtained from the
training procedure, which makes the algorithm very flexible with regards to
what data manifolds it can learn. We test its generalization performance and
speed on a range of benchmark datasets and detail in which settings it
outperforms the state of the art. Empirically we find that training time scales
as O(DNlog(N)) and testing as O(Dlog(N)), where D is the dimensionality and N
the amount of data,
| no_new_dataset | 0.946151 |
1505.02973 | Konstantinos Tserpes | Evangelos Psomakelis, Konstantinos Tserpes, Dimosthenis
Anagnostopoulos, Theodora Varvarigou | Comparing methods for Twitter Sentiment Analysis | 5 pages, 1 figure, 6th Conference on Knowledge Discovery and
Information Retrieval 2014, Rome, Italy | null | null | null | cs.CL cs.IR cs.SI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This work extends the set of works which deal with the popular problem of
sentiment analysis in Twitter. It investigates the most popular document
("tweet") representation methods which feed sentiment evaluation mechanisms. In
particular, we study the bag-of-words, n-grams and n-gram graphs approaches and
for each of them we evaluate the performance of a lexicon-based and 7
learning-based classification algorithms (namely SVM, Na\"ive Bayesian
Networks, Logistic Regression, Multilayer Perceptrons, Best-First Trees,
Functional Trees and C4.5) as well as their combinations, using a set of 4451
manually annotated tweets. The results demonstrate the superiority of
learning-based methods and in particular of n-gram graphs approaches for
predicting the sentiment of tweets. They also show that the combinatory
approach has impressive effects on n-grams, raising the confidence up to 83.15%
on the 5-Grams, using majority vote and a balanced dataset (equal number of
positive, negative and neutral tweets for training). In the n-gram graph cases
the improvement was small to none, reaching 94.52% on the 4-gram graphs, using
Orthodromic distance and a threshold of 0.001.
| [
{
"version": "v1",
"created": "Tue, 12 May 2015 12:05:19 GMT"
}
] | 2015-05-14T00:00:00 | [
[
"Psomakelis",
"Evangelos",
""
],
[
"Tserpes",
"Konstantinos",
""
],
[
"Anagnostopoulos",
"Dimosthenis",
""
],
[
"Varvarigou",
"Theodora",
""
]
] | TITLE: Comparing methods for Twitter Sentiment Analysis
ABSTRACT: This work extends the set of works which deal with the popular problem of
sentiment analysis in Twitter. It investigates the most popular document
("tweet") representation methods which feed sentiment evaluation mechanisms. In
particular, we study the bag-of-words, n-grams and n-gram graphs approaches and
for each of them we evaluate the performance of a lexicon-based and 7
learning-based classification algorithms (namely SVM, Na\"ive Bayesian
Networks, Logistic Regression, Multilayer Perceptrons, Best-First Trees,
Functional Trees and C4.5) as well as their combinations, using a set of 4451
manually annotated tweets. The results demonstrate the superiority of
learning-based methods and in particular of n-gram graphs approaches for
predicting the sentiment of tweets. They also show that the combinatory
approach has impressive effects on n-grams, raising the confidence up to 83.15%
on the 5-Grams, using majority vote and a balanced dataset (equal number of
positive, negative and neutral tweets for training). In the n-gram graph cases
the improvement was small to none, reaching 94.52% on the 4-gram graphs, using
Orthodromic distance and a threshold of 0.001.
| no_new_dataset | 0.946001 |
1505.03229 | Ikuro Sato | Ikuro Sato, Hiroki Nishimura, Kensuke Yokoi | APAC: Augmented PAttern Classification with Neural Networks | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Deep neural networks have been exhibiting splendid accuracies in many of
visual pattern classification problems. Many of the state-of-the-art methods
employ a technique known as data augmentation at the training stage. This paper
addresses an issue of decision rule for classifiers trained with augmented
data. Our method is named as APAC: the Augmented PAttern Classification, which
is a way of classification using the optimal decision rule for augmented data
learning. Discussion of methods of data augmentation is not our primary focus.
We show clear evidences that APAC gives far better generalization performance
than the traditional way of class prediction in several experiments. Our
convolutional neural network model with APAC achieved a state-of-the-art
accuracy on the MNIST dataset among non-ensemble classifiers. Even our
multilayer perceptron model beats some of the convolutional models with
recently invented stochastic regularization techniques on the CIFAR-10 dataset.
| [
{
"version": "v1",
"created": "Wed, 13 May 2015 03:33:29 GMT"
}
] | 2015-05-14T00:00:00 | [
[
"Sato",
"Ikuro",
""
],
[
"Nishimura",
"Hiroki",
""
],
[
"Yokoi",
"Kensuke",
""
]
] | TITLE: APAC: Augmented PAttern Classification with Neural Networks
ABSTRACT: Deep neural networks have been exhibiting splendid accuracies in many of
visual pattern classification problems. Many of the state-of-the-art methods
employ a technique known as data augmentation at the training stage. This paper
addresses an issue of decision rule for classifiers trained with augmented
data. Our method is named as APAC: the Augmented PAttern Classification, which
is a way of classification using the optimal decision rule for augmented data
learning. Discussion of methods of data augmentation is not our primary focus.
We show clear evidences that APAC gives far better generalization performance
than the traditional way of class prediction in several experiments. Our
convolutional neural network model with APAC achieved a state-of-the-art
accuracy on the MNIST dataset among non-ensemble classifiers. Even our
multilayer perceptron model beats some of the convolutional models with
recently invented stochastic regularization techniques on the CIFAR-10 dataset.
| no_new_dataset | 0.947672 |
1505.03236 | Jensi | R. Jensi and G. Wiselin Jiji | Hybrid data clustering approach using K-Means and Flower Pollination
Algorithm | 11 pages, Journal. Advanced Computational Intelligence: An
International Journal (ACII), Vol.2, No.2, April 2015 | null | null | null | cs.LG cs.IR cs.NE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Data clustering is a technique for clustering set of objects into known
number of groups. Several approaches are widely applied to data clustering so
that objects within the clusters are similar and objects in different clusters
are far away from each other. K-Means, is one of the familiar center based
clustering algorithms since implementation is very easy and fast convergence.
However, K-Means algorithm suffers from initialization, hence trapped in local
optima. Flower Pollination Algorithm (FPA) is the global optimization
technique, which avoids trapping in local optimum solution. In this paper, a
novel hybrid data clustering approach using Flower Pollination Algorithm and
K-Means (FPAKM) is proposed. The proposed algorithm results are compared with
K-Means and FPA on eight datasets. From the experimental results, FPAKM is
better than FPA and K-Means.
| [
{
"version": "v1",
"created": "Wed, 13 May 2015 04:24:50 GMT"
}
] | 2015-05-14T00:00:00 | [
[
"Jensi",
"R.",
""
],
[
"Jiji",
"G. Wiselin",
""
]
] | TITLE: Hybrid data clustering approach using K-Means and Flower Pollination
Algorithm
ABSTRACT: Data clustering is a technique for clustering set of objects into known
number of groups. Several approaches are widely applied to data clustering so
that objects within the clusters are similar and objects in different clusters
are far away from each other. K-Means, is one of the familiar center based
clustering algorithms since implementation is very easy and fast convergence.
However, K-Means algorithm suffers from initialization, hence trapped in local
optima. Flower Pollination Algorithm (FPA) is the global optimization
technique, which avoids trapping in local optimum solution. In this paper, a
novel hybrid data clustering approach using Flower Pollination Algorithm and
K-Means (FPAKM) is proposed. The proposed algorithm results are compared with
K-Means and FPA on eight datasets. From the experimental results, FPAKM is
better than FPA and K-Means.
| no_new_dataset | 0.951006 |
1505.03329 | Jelena Milosevic | Jelena Milosevic, Alberto Ferrante, Miroslaw Malek | A general practitioner or a specialist for your infected smartphone? | 2 pages poster, IEEE Symposium on Security and Privacy, San Jose,
USA, May 2015 | null | null | null | cs.CR | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | With explosive growth in the number of mobile devices, the mobile malware is
rapidly spreading as well, and the number of encountered malware families is
increasing. Existing solutions, which are mainly based on one malware detector
running on the phone or in the cloud, are no longer effective. Main problem
lies in the fact that it might be impossible to create a unique mobile malware
detector that would be able to detect different malware families with high
accuracy, being at the same time lightweight enough not to drain battery
quickly and fast enough to give results of detection promptly. The proposed
approach to mobile malware detection is analogous to general practitioner
versus specialist approach to dealing with a medical problem. Similarly to a
general practitioner that, based on indicative symptoms identifies potential
illnesses and sends the patient to an appropriate specialist, our detection
system distinguishes among symptoms representing different malware families
and, once the symptoms are detected, it triggers specific analyses. A system
monitoring application operates in the same way as a general practitioner. It
is able to distinguish between different symptoms and trigger appropriate
detection mechanisms. As an analogy to different specialists, an ensemble of
detectors, each of which specifically trained for a particular malware family,
is used. The main challenge of the approach is to define representative
symptoms of different malware families and train detectors accordingly to them.
The main goal of the poster is to foster discussion on the most representative
symptoms of different malware families and to discuss initial results in this
area obtained by using Malware Genome project dataset.
| [
{
"version": "v1",
"created": "Wed, 13 May 2015 11:09:52 GMT"
}
] | 2015-05-14T00:00:00 | [
[
"Milosevic",
"Jelena",
""
],
[
"Ferrante",
"Alberto",
""
],
[
"Malek",
"Miroslaw",
""
]
] | TITLE: A general practitioner or a specialist for your infected smartphone?
ABSTRACT: With explosive growth in the number of mobile devices, the mobile malware is
rapidly spreading as well, and the number of encountered malware families is
increasing. Existing solutions, which are mainly based on one malware detector
running on the phone or in the cloud, are no longer effective. Main problem
lies in the fact that it might be impossible to create a unique mobile malware
detector that would be able to detect different malware families with high
accuracy, being at the same time lightweight enough not to drain battery
quickly and fast enough to give results of detection promptly. The proposed
approach to mobile malware detection is analogous to general practitioner
versus specialist approach to dealing with a medical problem. Similarly to a
general practitioner that, based on indicative symptoms identifies potential
illnesses and sends the patient to an appropriate specialist, our detection
system distinguishes among symptoms representing different malware families
and, once the symptoms are detected, it triggers specific analyses. A system
monitoring application operates in the same way as a general practitioner. It
is able to distinguish between different symptoms and trigger appropriate
detection mechanisms. As an analogy to different specialists, an ensemble of
detectors, each of which specifically trained for a particular malware family,
is used. The main challenge of the approach is to define representative
symptoms of different malware families and train detectors accordingly to them.
The main goal of the poster is to foster discussion on the most representative
symptoms of different malware families and to discuss initial results in this
area obtained by using Malware Genome project dataset.
| no_new_dataset | 0.922412 |
0811.2802 | Perivolaropoulos Leandros | L. Perivolaropoulos and A. Shafieloo | Bright High z SnIa: A Challenge for LCDM? | 8 pages, 7 figures. Extended analysis, added two figures, results
unchanged. The data and the mathematica v.5.2 files used for the production
of the figures may be downloaded from http://leandros.physics.uoi.gr/bnd.zip.
Accepted in Physical Review D (to appear) | Phys.Rev.D79:123502,2009 | 10.1103/PhysRevD.79.123502 | null | astro-ph gr-qc physics.data-an | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | It has recently been pointed out by Kowalski et. al. (arxiv:0804.4142) that
there is `an unexpected brightness of the SnIa data at z>1'. We quantify this
statement by constructing a new statistic which is applicable directly on the
Type Ia Supernova (SnIa) distance moduli. This statistic is designed to pick up
systematic brightness trends of SnIa datapoints with respect to a best fit
cosmological model at high redshifts. It is based on binning the normalized
differences between the SnIa distance moduli and the corresponding best fit
values in the context of a specific cosmological model (eg LCDM). We then focus
on the highest redshift bin and extend its size towards lower redshifts until
the Binned Normalized Difference (BND) changes sign (crosses 0) at a redshift
z_c (bin size N_c). The bin size N_c of this crossing (the statistical
variable) is then compared with the corresponding crossing bin size N_{mc} for
Monte Carlo data realizations based on the best fit model. We find that the
crossing bin size N_c obtained from the Union08 and Gold06 data with respect to
the best fit LCDM model is anomalously large compared to N_{mc} of the
corresponding Monte Carlo datasets obtained from the best fit LCDM in each
case. In particular, only 2.2% of the Monte Carlo LCDM datasets are consistent
with the Gold06 value of N_c while the corresponding probability for the
Union08 value of N_c is 5.3%. Thus, according to this statistic, the
probability that the high redshift brightness bias of the Union08 and Gold06
datasets is realized in the context of a (w_0,w_1)=(-1,0) model (LCDM
cosmology) is less than 6%. The corresponding realization probability in the
context of a (w_0,w_1)=(-1.4,2) model is more than 30% for both the Union08 and
the Gold06 datasets.
| [
{
"version": "v1",
"created": "Mon, 17 Nov 2008 21:03:50 GMT"
},
{
"version": "v2",
"created": "Mon, 1 Jun 2009 15:24:13 GMT"
}
] | 2015-05-13T00:00:00 | [
[
"Perivolaropoulos",
"L.",
""
],
[
"Shafieloo",
"A.",
""
]
] | TITLE: Bright High z SnIa: A Challenge for LCDM?
ABSTRACT: It has recently been pointed out by Kowalski et. al. (arxiv:0804.4142) that
there is `an unexpected brightness of the SnIa data at z>1'. We quantify this
statement by constructing a new statistic which is applicable directly on the
Type Ia Supernova (SnIa) distance moduli. This statistic is designed to pick up
systematic brightness trends of SnIa datapoints with respect to a best fit
cosmological model at high redshifts. It is based on binning the normalized
differences between the SnIa distance moduli and the corresponding best fit
values in the context of a specific cosmological model (eg LCDM). We then focus
on the highest redshift bin and extend its size towards lower redshifts until
the Binned Normalized Difference (BND) changes sign (crosses 0) at a redshift
z_c (bin size N_c). The bin size N_c of this crossing (the statistical
variable) is then compared with the corresponding crossing bin size N_{mc} for
Monte Carlo data realizations based on the best fit model. We find that the
crossing bin size N_c obtained from the Union08 and Gold06 data with respect to
the best fit LCDM model is anomalously large compared to N_{mc} of the
corresponding Monte Carlo datasets obtained from the best fit LCDM in each
case. In particular, only 2.2% of the Monte Carlo LCDM datasets are consistent
with the Gold06 value of N_c while the corresponding probability for the
Union08 value of N_c is 5.3%. Thus, according to this statistic, the
probability that the high redshift brightness bias of the Union08 and Gold06
datasets is realized in the context of a (w_0,w_1)=(-1,0) model (LCDM
cosmology) is less than 6%. The corresponding realization probability in the
context of a (w_0,w_1)=(-1.4,2) model is more than 30% for both the Union08 and
the Gold06 datasets.
| no_new_dataset | 0.956145 |
0812.0743 | Qiang Li | Qiang Li, Yan He, Jing-ping Jiang | A Novel Clustering Algorithm Based on Quantum Games | 19 pages, 5 figures, 5 tables | 2009 J. Phys. A: Math. Theor. 42 445303 | 10.1088/1751-8113/42/44/445303 | null | cs.LG cs.AI cs.CV cs.GT cs.MA cs.NE quant-ph | http://creativecommons.org/licenses/by-nc-sa/3.0/ | Enormous successes have been made by quantum algorithms during the last
decade. In this paper, we combine the quantum game with the problem of data
clustering, and then develop a quantum-game-based clustering algorithm, in
which data points in a dataset are considered as players who can make decisions
and implement quantum strategies in quantum games. After each round of a
quantum game, each player's expected payoff is calculated. Later, he uses a
link-removing-and-rewiring (LRR) function to change his neighbors and adjust
the strength of links connecting to them in order to maximize his payoff.
Further, algorithms are discussed and analyzed in two cases of strategies, two
payoff matrixes and two LRR functions. Consequently, the simulation results
have demonstrated that data points in datasets are clustered reasonably and
efficiently, and the clustering algorithms have fast rates of convergence.
Moreover, the comparison with other algorithms also provides an indication of
the effectiveness of the proposed approach.
| [
{
"version": "v1",
"created": "Wed, 3 Dec 2008 15:46:03 GMT"
},
{
"version": "v2",
"created": "Sat, 10 Oct 2009 09:10:36 GMT"
}
] | 2015-05-13T00:00:00 | [
[
"Li",
"Qiang",
""
],
[
"He",
"Yan",
""
],
[
"Jiang",
"Jing-ping",
""
]
] | TITLE: A Novel Clustering Algorithm Based on Quantum Games
ABSTRACT: Enormous successes have been made by quantum algorithms during the last
decade. In this paper, we combine the quantum game with the problem of data
clustering, and then develop a quantum-game-based clustering algorithm, in
which data points in a dataset are considered as players who can make decisions
and implement quantum strategies in quantum games. After each round of a
quantum game, each player's expected payoff is calculated. Later, he uses a
link-removing-and-rewiring (LRR) function to change his neighbors and adjust
the strength of links connecting to them in order to maximize his payoff.
Further, algorithms are discussed and analyzed in two cases of strategies, two
payoff matrixes and two LRR functions. Consequently, the simulation results
have demonstrated that data points in datasets are clustered reasonably and
efficiently, and the clustering algorithms have fast rates of convergence.
Moreover, the comparison with other algorithms also provides an indication of
the effectiveness of the proposed approach.
| no_new_dataset | 0.950595 |
0902.4478 | David Pretty | D. G. Pretty and B. D. Blackwell | A data mining algorithm for automated characterisation of fluctuations
in multichannel timeseries | 14 pages, 8 figures | null | 10.1016/j.cpc.2009.05.003 | null | physics.data-an physics.plasm-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We present a data mining technique for the analysis of multichannel
oscillatory timeseries data and show an application using poloidal arrays of
magnetic sensors installed in the H-1 heliac. The procedure is highly
automated, and scales well to large datasets. The timeseries data is split into
short time segments to provide time resolution, and each segment is represented
by a singular value decomposition (SVD). By comparing power spectra of the
temporal singular vectors, singular values are grouped into subsets which
define fluctuation structures. Thresholds for the normalised energy of the
fluctuation structure and the normalised entropy of the SVD are used to filter
the dataset. We assume that distinct classes of fluctuations are localised in
the space of phase differences between each pair of nearest neighbour channels.
An expectation maximisation clustering algorithm is used to locate the distinct
classes of fluctuations, and a cluster tree mapping is used to visualise the
results.
| [
{
"version": "v1",
"created": "Wed, 25 Feb 2009 22:48:09 GMT"
}
] | 2015-05-13T00:00:00 | [
[
"Pretty",
"D. G.",
""
],
[
"Blackwell",
"B. D.",
""
]
] | TITLE: A data mining algorithm for automated characterisation of fluctuations
in multichannel timeseries
ABSTRACT: We present a data mining technique for the analysis of multichannel
oscillatory timeseries data and show an application using poloidal arrays of
magnetic sensors installed in the H-1 heliac. The procedure is highly
automated, and scales well to large datasets. The timeseries data is split into
short time segments to provide time resolution, and each segment is represented
by a singular value decomposition (SVD). By comparing power spectra of the
temporal singular vectors, singular values are grouped into subsets which
define fluctuation structures. Thresholds for the normalised energy of the
fluctuation structure and the normalised entropy of the SVD are used to filter
the dataset. We assume that distinct classes of fluctuations are localised in
the space of phase differences between each pair of nearest neighbour channels.
An expectation maximisation clustering algorithm is used to locate the distinct
classes of fluctuations, and a cluster tree mapping is used to visualise the
results.
| no_new_dataset | 0.949902 |
0906.2914 | Michal Zerola | Michal Zerola, J\'er\^ome Lauret, Roman Bart\'ak and Michal
\v{S}umbera | Efficient Multi-site Data Movement Using Constraint Programming for Data
Hungry Science | To appear in proceedings of Computing in High Energy and Nuclear
Physics 2009 | null | 10.1088/1742-6596/219/6/062069 | null | cs.DC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | For the past decade, HENP experiments have been heading towards a distributed
computing model in an effort to concurrently process tasks over enormous data
sets that have been increasing in size as a function of time. In order to
optimize all available resources (geographically spread) and minimize the
processing time, it is necessary to face also the question of efficient data
transfers and placements. A key question is whether the time penalty for moving
the data to the computational resources is worth the presumed gain. Onward to
the truly distributed task scheduling we present the technique using a
Constraint Programming (CP) approach. The CP technique schedules data transfers
from multiple resources considering all available paths of diverse
characteristic (capacity, sharing and storage) having minimum user's waiting
time as an objective. We introduce a model for planning data transfers to a
single destination (data transfer) as well as its extension for an optimal data
set spreading strategy (data placement). Several enhancements for a solver of
the CP model will be shown, leading to a faster schedule computation time using
symmetry breaking, branch cutting, well studied principles from job-shop
scheduling field and several heuristics. Finally, we will present the design
and implementation of a corner-stone application aimed at moving datasets
according to the schedule. Results will include comparison of performance and
trade-off between CP techniques and a Peer-2-Peer model from simulation
framework as well as the real case scenario taken from a practical usage of a
CP scheduler.
| [
{
"version": "v1",
"created": "Tue, 16 Jun 2009 12:33:25 GMT"
},
{
"version": "v2",
"created": "Thu, 18 Jun 2009 19:20:05 GMT"
}
] | 2015-05-13T00:00:00 | [
[
"Zerola",
"Michal",
""
],
[
"Lauret",
"Jérôme",
""
],
[
"Barták",
"Roman",
""
],
[
"Šumbera",
"Michal",
""
]
] | TITLE: Efficient Multi-site Data Movement Using Constraint Programming for Data
Hungry Science
ABSTRACT: For the past decade, HENP experiments have been heading towards a distributed
computing model in an effort to concurrently process tasks over enormous data
sets that have been increasing in size as a function of time. In order to
optimize all available resources (geographically spread) and minimize the
processing time, it is necessary to face also the question of efficient data
transfers and placements. A key question is whether the time penalty for moving
the data to the computational resources is worth the presumed gain. Onward to
the truly distributed task scheduling we present the technique using a
Constraint Programming (CP) approach. The CP technique schedules data transfers
from multiple resources considering all available paths of diverse
characteristic (capacity, sharing and storage) having minimum user's waiting
time as an objective. We introduce a model for planning data transfers to a
single destination (data transfer) as well as its extension for an optimal data
set spreading strategy (data placement). Several enhancements for a solver of
the CP model will be shown, leading to a faster schedule computation time using
symmetry breaking, branch cutting, well studied principles from job-shop
scheduling field and several heuristics. Finally, we will present the design
and implementation of a corner-stone application aimed at moving datasets
according to the schedule. Results will include comparison of performance and
trade-off between CP techniques and a Peer-2-Peer model from simulation
framework as well as the real case scenario taken from a practical usage of a
CP scheduler.
| no_new_dataset | 0.946695 |
1412.6071 | Benjamin Graham | Benjamin Graham | Fractional Max-Pooling | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Convolutional networks almost always incorporate some form of spatial
pooling, and very often it is alpha times alpha max-pooling with alpha=2.
Max-pooling act on the hidden layers of the network, reducing their size by an
integer multiplicative factor alpha. The amazing by-product of discarding 75%
of your data is that you build into the network a degree of invariance with
respect to translations and elastic distortions. However, if you simply
alternate convolutional layers with max-pooling layers, performance is limited
due to the rapid reduction in spatial size, and the disjoint nature of the
pooling regions. We have formulated a fractional version of max-pooling where
alpha is allowed to take non-integer values. Our version of max-pooling is
stochastic as there are lots of different ways of constructing suitable pooling
regions. We find that our form of fractional max-pooling reduces overfitting on
a variety of datasets: for instance, we improve on the state-of-the art for
CIFAR-100 without even using dropout.
| [
{
"version": "v1",
"created": "Thu, 18 Dec 2014 20:45:11 GMT"
},
{
"version": "v2",
"created": "Mon, 22 Dec 2014 11:06:35 GMT"
},
{
"version": "v3",
"created": "Mon, 2 Mar 2015 20:06:22 GMT"
},
{
"version": "v4",
"created": "Tue, 12 May 2015 06:36:11 GMT"
}
] | 2015-05-13T00:00:00 | [
[
"Graham",
"Benjamin",
""
]
] | TITLE: Fractional Max-Pooling
ABSTRACT: Convolutional networks almost always incorporate some form of spatial
pooling, and very often it is alpha times alpha max-pooling with alpha=2.
Max-pooling act on the hidden layers of the network, reducing their size by an
integer multiplicative factor alpha. The amazing by-product of discarding 75%
of your data is that you build into the network a degree of invariance with
respect to translations and elastic distortions. However, if you simply
alternate convolutional layers with max-pooling layers, performance is limited
due to the rapid reduction in spatial size, and the disjoint nature of the
pooling regions. We have formulated a fractional version of max-pooling where
alpha is allowed to take non-integer values. Our version of max-pooling is
stochastic as there are lots of different ways of constructing suitable pooling
regions. We find that our form of fractional max-pooling reduces overfitting on
a variety of datasets: for instance, we improve on the state-of-the art for
CIFAR-100 without even using dropout.
| no_new_dataset | 0.947088 |
1412.6857 | Tyng-Luh Liu | Jyh-Jing Hwang and Tyng-Luh Liu | Contour Detection Using Cost-Sensitive Convolutional Neural Networks | 9 pages, 3 figures | null | null | null | cs.CV cs.LG cs.NE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We address the problem of contour detection via per-pixel classifications of
edge point. To facilitate the process, the proposed approach leverages with
DenseNet, an efficient implementation of multiscale convolutional neural
networks (CNNs), to extract an informative feature vector for each pixel and
uses an SVM classifier to accomplish contour detection. The main challenge lies
in adapting a pre-trained per-image CNN model for yielding per-pixel image
features. We propose to base on the DenseNet architecture to achieve pixelwise
fine-tuning and then consider a cost-sensitive strategy to further improve the
learning with a small dataset of edge and non-edge image patches. In the
experiment of contour detection, we look into the effectiveness of combining
per-pixel features from different CNN layers and obtain comparable performances
to the state-of-the-art on BSDS500.
| [
{
"version": "v1",
"created": "Mon, 22 Dec 2014 01:16:50 GMT"
},
{
"version": "v2",
"created": "Wed, 24 Dec 2014 14:37:27 GMT"
},
{
"version": "v3",
"created": "Thu, 15 Jan 2015 15:01:16 GMT"
},
{
"version": "v4",
"created": "Sat, 28 Feb 2015 07:37:54 GMT"
},
{
"version": "v5",
"created": "Tue, 12 May 2015 08:42:42 GMT"
}
] | 2015-05-13T00:00:00 | [
[
"Hwang",
"Jyh-Jing",
""
],
[
"Liu",
"Tyng-Luh",
""
]
] | TITLE: Contour Detection Using Cost-Sensitive Convolutional Neural Networks
ABSTRACT: We address the problem of contour detection via per-pixel classifications of
edge point. To facilitate the process, the proposed approach leverages with
DenseNet, an efficient implementation of multiscale convolutional neural
networks (CNNs), to extract an informative feature vector for each pixel and
uses an SVM classifier to accomplish contour detection. The main challenge lies
in adapting a pre-trained per-image CNN model for yielding per-pixel image
features. We propose to base on the DenseNet architecture to achieve pixelwise
fine-tuning and then consider a cost-sensitive strategy to further improve the
learning with a small dataset of edge and non-edge image patches. In the
experiment of contour detection, we look into the effectiveness of combining
per-pixel features from different CNN layers and obtain comparable performances
to the state-of-the-art on BSDS500.
| no_new_dataset | 0.946349 |
1504.06603 | Dmytro Mishkin | Dmytro Mishkin and Jiri Matas and Michal Perdoch and Karel Lenc | WxBS: Wide Baseline Stereo Generalizations | Descriptor and detector evaluation expanded | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We have presented a new problem -- the wide multiple baseline stereo (WxBS)
-- which considers matching of images that simultaneously differ in more than
one image acquisition factor such as viewpoint, illumination, sensor type or
where object appearance changes significantly, e.g. over time. A new dataset
with the ground truth for evaluation of matching algorithms has been introduced
and will be made public.
We have extensively tested a large set of popular and recent detectors and
descriptors and show than the combination of RootSIFT and HalfRootSIFT as
descriptors with MSER and Hessian-Affine detectors works best for many
different nuisance factors. We show that simple adaptive thresholding improves
Hessian-Affine, DoG, MSER (and possibly other) detectors and allows to use them
on infrared and low contrast images.
A novel matching algorithm for addressing the WxBS problem has been
introduced. We have shown experimentally that the WxBS-M matcher dominantes the
state-of-the-art methods both on both the new and existing datasets.
| [
{
"version": "v1",
"created": "Fri, 24 Apr 2015 19:19:04 GMT"
},
{
"version": "v2",
"created": "Tue, 12 May 2015 14:42:53 GMT"
}
] | 2015-05-13T00:00:00 | [
[
"Mishkin",
"Dmytro",
""
],
[
"Matas",
"Jiri",
""
],
[
"Perdoch",
"Michal",
""
],
[
"Lenc",
"Karel",
""
]
] | TITLE: WxBS: Wide Baseline Stereo Generalizations
ABSTRACT: We have presented a new problem -- the wide multiple baseline stereo (WxBS)
-- which considers matching of images that simultaneously differ in more than
one image acquisition factor such as viewpoint, illumination, sensor type or
where object appearance changes significantly, e.g. over time. A new dataset
with the ground truth for evaluation of matching algorithms has been introduced
and will be made public.
We have extensively tested a large set of popular and recent detectors and
descriptors and show than the combination of RootSIFT and HalfRootSIFT as
descriptors with MSER and Hessian-Affine detectors works best for many
different nuisance factors. We show that simple adaptive thresholding improves
Hessian-Affine, DoG, MSER (and possibly other) detectors and allows to use them
on infrared and low contrast images.
A novel matching algorithm for addressing the WxBS problem has been
introduced. We have shown experimentally that the WxBS-M matcher dominantes the
state-of-the-art methods both on both the new and existing datasets.
| new_dataset | 0.958226 |
1505.02982 | Baoguang Shi | Baoguang Shi, Cong Yao, Chengquan Zhang, Xiaowei Guo, Feiyue Huang,
Xiang Bai | Automatic Script Identification in the Wild | 5 pages, 7 figures, submitted to ICDAR 2015 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | With the rapid increase of transnational communication and cooperation,
people frequently encounter multilingual scenarios in various situations. In
this paper, we are concerned with a relatively new problem: script
identification at word or line levels in natural scenes. A large-scale dataset
with a great quantity of natural images and 10 types of widely used languages
is constructed and released. In allusion to the challenges in script
identification in real-world scenarios, a deep learning based algorithm is
proposed. The experiments on the proposed dataset demonstrate that our
algorithm achieves superior performance, compared with conventional image
classification methods, such as the original CNN architecture and LLC.
| [
{
"version": "v1",
"created": "Tue, 12 May 2015 12:38:30 GMT"
}
] | 2015-05-13T00:00:00 | [
[
"Shi",
"Baoguang",
""
],
[
"Yao",
"Cong",
""
],
[
"Zhang",
"Chengquan",
""
],
[
"Guo",
"Xiaowei",
""
],
[
"Huang",
"Feiyue",
""
],
[
"Bai",
"Xiang",
""
]
] | TITLE: Automatic Script Identification in the Wild
ABSTRACT: With the rapid increase of transnational communication and cooperation,
people frequently encounter multilingual scenarios in various situations. In
this paper, we are concerned with a relatively new problem: script
identification at word or line levels in natural scenes. A large-scale dataset
with a great quantity of natural images and 10 types of widely used languages
is constructed and released. In allusion to the challenges in script
identification in real-world scenarios, a deep learning based algorithm is
proposed. The experiments on the proposed dataset demonstrate that our
algorithm achieves superior performance, compared with conventional image
classification methods, such as the original CNN architecture and LLC.
| new_dataset | 0.955899 |
1505.03008 | Lovro \v{S}ubelj | Dalibor Fiala, Lovro \v{S}ubelj, Slavko \v{Z}itnik, Marko Bajec | Do PageRank-based author rankings outperform simple citation counts? | 28 pages, 5 figures, 6 tables | J. Infometr. 9(2), 334-348 (2015) | 10.1016/j.joi.2015.02.008 | null | cs.DL cs.SI physics.soc-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The basic indicators of a researcher's productivity and impact are still the
number of publications and their citation counts. These metrics are clear,
straightforward, and easy to obtain. When a ranking of scholars is needed, for
instance in grant, award, or promotion procedures, their use is the fastest and
cheapest way of prioritizing some scientists over others. However, due to their
nature, there is a danger of oversimplifying scientific achievements.
Therefore, many other indicators have been proposed including the usage of the
PageRank algorithm known for the ranking of webpages and its modifications
suited to citation networks. Nevertheless, this recursive method is
computationally expensive and even if it has the advantage of favouring
prestige over popularity, its application should be well justified,
particularly when compared to the standard citation counts. In this study, we
analyze three large datasets of computer science papers in the categories of
artificial intelligence, software engineering, and theory and methods and apply
12 different ranking methods to the citation networks of authors. We compare
the resulting rankings with self-compiled lists of outstanding researchers
selected as frequent editorial board members of prestigious journals in the
field and conclude that there is no evidence of PageRank-based methods
outperforming simple citation counts.
| [
{
"version": "v1",
"created": "Tue, 12 May 2015 13:59:30 GMT"
}
] | 2015-05-13T00:00:00 | [
[
"Fiala",
"Dalibor",
""
],
[
"Šubelj",
"Lovro",
""
],
[
"Žitnik",
"Slavko",
""
],
[
"Bajec",
"Marko",
""
]
] | TITLE: Do PageRank-based author rankings outperform simple citation counts?
ABSTRACT: The basic indicators of a researcher's productivity and impact are still the
number of publications and their citation counts. These metrics are clear,
straightforward, and easy to obtain. When a ranking of scholars is needed, for
instance in grant, award, or promotion procedures, their use is the fastest and
cheapest way of prioritizing some scientists over others. However, due to their
nature, there is a danger of oversimplifying scientific achievements.
Therefore, many other indicators have been proposed including the usage of the
PageRank algorithm known for the ranking of webpages and its modifications
suited to citation networks. Nevertheless, this recursive method is
computationally expensive and even if it has the advantage of favouring
prestige over popularity, its application should be well justified,
particularly when compared to the standard citation counts. In this study, we
analyze three large datasets of computer science papers in the categories of
artificial intelligence, software engineering, and theory and methods and apply
12 different ranking methods to the citation networks of authors. We compare
the resulting rankings with self-compiled lists of outstanding researchers
selected as frequent editorial board members of prestigious journals in the
field and conclude that there is no evidence of PageRank-based methods
outperforming simple citation counts.
| no_new_dataset | 0.941277 |
1505.03093 | Jose Luis Alves | Manuel Pinheiro and J.L. Alves | A new Level-set based Protocol for Accurate Bone Segmentation from CT
Imaging | 11 pages, 24 figures | null | null | null | physics.med-ph cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this work it is proposed a medical image segmentation pipeline for
accurate bone segmentation from CT imaging. It is a two-step methodology, with
a pre-segmentation step and a segmentation refinement step. First, the user
performs a rough segmenting of the desired region of interest. Next, a fully
automatic refinement step is applied to the pre-segmented data. The automatic
segmentation refinement is composed by several sub-stpng, namely image
deconvolution, image cropping and interpolation. The user-defined
pre-segmentation is then refined over the deconvolved, cropped, and up-sampled
version of the image. The algorithm is applied in the segmentation of CT images
of a composite femur bone, reconstructed with different reconstruction
protocols. Segmentation outcomes are validated against a gold standard model
obtained with coordinate measuring machine Nikon Metris LK V20 with a digital
line scanner LC60-D that guarantees an accuracy of 28 $\mu m$. High sub-pixel
accuracy models were obtained for all tested Datasets. The algorithm is able to
produce high quality segmentation of the composite femur regardless of the
surface meshing strategy used.
| [
{
"version": "v1",
"created": "Tue, 12 May 2015 17:33:44 GMT"
}
] | 2015-05-13T00:00:00 | [
[
"Pinheiro",
"Manuel",
""
],
[
"Alves",
"J. L.",
""
]
] | TITLE: A new Level-set based Protocol for Accurate Bone Segmentation from CT
Imaging
ABSTRACT: In this work it is proposed a medical image segmentation pipeline for
accurate bone segmentation from CT imaging. It is a two-step methodology, with
a pre-segmentation step and a segmentation refinement step. First, the user
performs a rough segmenting of the desired region of interest. Next, a fully
automatic refinement step is applied to the pre-segmented data. The automatic
segmentation refinement is composed by several sub-stpng, namely image
deconvolution, image cropping and interpolation. The user-defined
pre-segmentation is then refined over the deconvolved, cropped, and up-sampled
version of the image. The algorithm is applied in the segmentation of CT images
of a composite femur bone, reconstructed with different reconstruction
protocols. Segmentation outcomes are validated against a gold standard model
obtained with coordinate measuring machine Nikon Metris LK V20 with a digital
line scanner LC60-D that guarantees an accuracy of 28 $\mu m$. High sub-pixel
accuracy models were obtained for all tested Datasets. The algorithm is able to
produce high quality segmentation of the composite femur regardless of the
surface meshing strategy used.
| no_new_dataset | 0.949106 |
1408.6221 | Amir Gholami | Amir Gholami, Andreas Mang, George Biros | An inverse problem formulation for parameter estimation of a reaction
diffusion model of low grade gliomas | J. Mat. Bio. (2015) | null | 10.1007/s00285-015-0888-x | null | math.NA physics.med-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We present a numerical scheme for solving a parameter estimation problem for
a model of low-grade glioma growth. Our goal is to estimate the spatial
distribution of tumor concentration, as well as the magnitude of anisotropic
tumor diffusion. We use a constrained optimization formulation with a
reaction-diffusion model that results in a system of nonlinear partial
differential equations (PDEs). In our formulation, we estimate the parameters
using partially observed, noisy tumor concentration data at two different time
instances, along with white matter fiber directions derived from diffusion
tensor imaging (DTI). The optimization problem is solved with a Gauss-Newton
reduced space algorithm. We present the formulation and outline the numerical
algorithms for solving the resulting equations. We test the method using a
synthetic dataset and compute the reconstruction error for different noise
levels and detection thresholds for monofocal and multifocal test cases.
| [
{
"version": "v1",
"created": "Tue, 26 Aug 2014 19:49:31 GMT"
},
{
"version": "v2",
"created": "Wed, 27 Aug 2014 19:14:52 GMT"
},
{
"version": "v3",
"created": "Mon, 11 May 2015 18:13:50 GMT"
}
] | 2015-05-12T00:00:00 | [
[
"Gholami",
"Amir",
""
],
[
"Mang",
"Andreas",
""
],
[
"Biros",
"George",
""
]
] | TITLE: An inverse problem formulation for parameter estimation of a reaction
diffusion model of low grade gliomas
ABSTRACT: We present a numerical scheme for solving a parameter estimation problem for
a model of low-grade glioma growth. Our goal is to estimate the spatial
distribution of tumor concentration, as well as the magnitude of anisotropic
tumor diffusion. We use a constrained optimization formulation with a
reaction-diffusion model that results in a system of nonlinear partial
differential equations (PDEs). In our formulation, we estimate the parameters
using partially observed, noisy tumor concentration data at two different time
instances, along with white matter fiber directions derived from diffusion
tensor imaging (DTI). The optimization problem is solved with a Gauss-Newton
reduced space algorithm. We present the formulation and outline the numerical
algorithms for solving the resulting equations. We test the method using a
synthetic dataset and compute the reconstruction error for different noise
levels and detection thresholds for monofocal and multifocal test cases.
| no_new_dataset | 0.947332 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.