Split (1)

train · 174k rows

Search is not available for this dataset

id string	submitter string	authors string	title string	comments string	journal-ref string	doi string	report-no string	categories string	license string	abstract string	versions list	update_date timestamp[s]	authors_parsed sequence	prompt string
1210.5338	Cyril Furtlehner	Cyril Furtlehner, Yufei Han, Jean-Marc Lasgouttes and Victorin Martin	Pairwise MRF Calibration by Perturbation of the Bethe Reference Point	54 pages, 8 figure. section 5 and refs added in V2	null	null	Inria RR-8059	cond-mat.dis-nn cond-mat.stat-mech cs.LG stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	We investigate different ways of generating approximate solutions to the pairwise Markov random field (MRF) selection problem. We focus mainly on the inverse Ising problem, but discuss also the somewhat related inverse Gaussian problem because both types of MRF are suitable for inference tasks with the belief propagation algorithm (BP) under certain conditions. Our approach consists in to take a Bethe mean-field solution obtained with a maximum spanning tree (MST) of pairwise mutual information, referred to as the \emph{Bethe reference point}, for further perturbation procedures. We consider three different ways following this idea: in the first one, we select and calibrate iteratively the optimal links to be added starting from the Bethe reference point; the second one is based on the observation that the natural gradient can be computed analytically at the Bethe point; in the third one, assuming no local field and using low temperature expansion we develop a dual loop joint model based on a well chosen fundamental cycle basis. We indeed identify a subclass of planar models, which we refer to as \emph{Bethe-dual graph models}, having possibly many loops, but characterized by a singly connected dual factor graph, for which the partition function and the linear response can be computed exactly in respectively O(N) and $O(N^2)$ operations, thanks to a dual weight propagation (DWP) message passing procedure that we set up. When restricted to this subclass of models, the inverse Ising problem being convex, becomes tractable at any temperature. Experimental tests on various datasets with refined $L_0$ or $L_1$ regularization procedures indicate that these approaches may be competitive and useful alternatives to existing ones.	[ { "version": "v1", "created": "Fri, 19 Oct 2012 08:08:55 GMT" }, { "version": "v2", "created": "Fri, 1 Feb 2013 17:32:44 GMT" } ]	2013-02-04T00:00:00	[ [ "Furtlehner", "Cyril", "" ], [ "Han", "Yufei", "" ], [ "Lasgouttes", "Jean-Marc", "" ], [ "Martin", "Victorin", "" ] ]	TITLE: Pairwise MRF Calibration by Perturbation of the Bethe Reference Point ABSTRACT: We investigate different ways of generating approximate solutions to the pairwise Markov random field (MRF) selection problem. We focus mainly on the inverse Ising problem, but discuss also the somewhat related inverse Gaussian problem because both types of MRF are suitable for inference tasks with the belief propagation algorithm (BP) under certain conditions. Our approach consists in to take a Bethe mean-field solution obtained with a maximum spanning tree (MST) of pairwise mutual information, referred to as the \emph{Bethe reference point}, for further perturbation procedures. We consider three different ways following this idea: in the first one, we select and calibrate iteratively the optimal links to be added starting from the Bethe reference point; the second one is based on the observation that the natural gradient can be computed analytically at the Bethe point; in the third one, assuming no local field and using low temperature expansion we develop a dual loop joint model based on a well chosen fundamental cycle basis. We indeed identify a subclass of planar models, which we refer to as \emph{Bethe-dual graph models}, having possibly many loops, but characterized by a singly connected dual factor graph, for which the partition function and the linear response can be computed exactly in respectively O(N) and $O(N^2)$ operations, thanks to a dual weight propagation (DWP) message passing procedure that we set up. When restricted to this subclass of models, the inverse Ising problem being convex, becomes tractable at any temperature. Experimental tests on various datasets with refined $L_0$ or $L_1$ regularization procedures indicate that these approaches may be competitive and useful alternatives to existing ones.
1111.5534	Lazaros Gallos	Lazaros K. Gallos, Diego Rybski, Fredrik Liljeros, Shlomo Havlin, Hernan A. Makse	How people interact in evolving online affiliation networks	10 pages, 8 figures	Phys. Rev. X 2, 031014 (2012)	null	null	physics.soc-ph cs.SI	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	The study of human interactions is of central importance for understanding the behavior of individuals, groups and societies. Here, we observe the formation and evolution of networks by monitoring the addition of all new links and we analyze quantitatively the tendencies used to create ties in these evolving online affiliation networks. We first show that an accurate estimation of these probabilistic tendencies can only be achieved by following the time evolution of the network. For example, actions that are attributed to the usual friend of a friend mechanism through a static snapshot of the network are overestimated by a factor of two. A detailed analysis of the dynamic network evolution shows that half of those triangles were generated through other mechanisms, in spite of the characteristic static pattern. We start by characterizing every single link when the tie was established in the network. This allows us to describe the probabilistic tendencies of tie formation and extract sociological conclusions as follows. The tendencies to add new links differ significantly from what we would expect if they were not affected by the individuals' structural position in the network, i.e., from random link formation. We also find significant differences in behavioral traits among individuals according to their degree of activity, gender, age, popularity and other attributes. For instance, in the particular datasets analyzed here, we find that women reciprocate connections three times as much as men and this difference increases with age. Men tend to connect with the most popular people more often than women across all ages. On the other hand, triangular ties tendencies are similar and independent of gender. Our findings can be useful to build models of realistic social network structures and discover the underlying laws that govern establishment of ties in evolving social networks.	[ { "version": "v1", "created": "Wed, 23 Nov 2011 16:04:06 GMT" } ]	2013-02-01T00:00:00	[ [ "Gallos", "Lazaros K.", "" ], [ "Rybski", "Diego", "" ], [ "Liljeros", "Fredrik", "" ], [ "Havlin", "Shlomo", "" ], [ "Makse", "Hernan A.", "" ] ]	TITLE: How people interact in evolving online affiliation networks ABSTRACT: The study of human interactions is of central importance for understanding the behavior of individuals, groups and societies. Here, we observe the formation and evolution of networks by monitoring the addition of all new links and we analyze quantitatively the tendencies used to create ties in these evolving online affiliation networks. We first show that an accurate estimation of these probabilistic tendencies can only be achieved by following the time evolution of the network. For example, actions that are attributed to the usual friend of a friend mechanism through a static snapshot of the network are overestimated by a factor of two. A detailed analysis of the dynamic network evolution shows that half of those triangles were generated through other mechanisms, in spite of the characteristic static pattern. We start by characterizing every single link when the tie was established in the network. This allows us to describe the probabilistic tendencies of tie formation and extract sociological conclusions as follows. The tendencies to add new links differ significantly from what we would expect if they were not affected by the individuals' structural position in the network, i.e., from random link formation. We also find significant differences in behavioral traits among individuals according to their degree of activity, gender, age, popularity and other attributes. For instance, in the particular datasets analyzed here, we find that women reciprocate connections three times as much as men and this difference increases with age. Men tend to connect with the most popular people more often than women across all ages. On the other hand, triangular ties tendencies are similar and independent of gender. Our findings can be useful to build models of realistic social network structures and discover the underlying laws that govern establishment of ties in evolving social networks.
1301.7363	John S. Breese	John S. Breese, David Heckerman, Carl Kadie	Empirical Analysis of Predictive Algorithms for Collaborative Filtering	Appears in Proceedings of the Fourteenth Conference on Uncertainty in Artificial Intelligence (UAI1998)	null	null	UAI-P-1998-PG-43-52	cs.IR cs.LG	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Collaborative filtering or recommender systems use a database about user preferences to predict additional topics or products a new user might like. In this paper we describe several algorithms designed for this task, including techniques based on correlation coefficients, vector-based similarity calculations, and statistical Bayesian methods. We compare the predictive accuracy of the various methods in a set of representative problem domains. We use two basic classes of evaluation metrics. The first characterizes accuracy over a set of individual predictions in terms of average absolute deviation. The second estimates the utility of a ranked list of suggested items. This metric uses an estimate of the probability that a user will see a recommendation in an ordered list. Experiments were run for datasets associated with 3 application areas, 4 experimental protocols, and the 2 evaluation metrics for the various algorithms. Results indicate that for a wide range of conditions, Bayesian networks with decision trees at each node and correlation methods outperform Bayesian-clustering and vector-similarity methods. Between correlation and Bayesian networks, the preferred method depends on the nature of the dataset, nature of the application (ranked versus one-by-one presentation), and the availability of votes with which to make predictions. Other considerations include the size of database, speed of predictions, and learning time.	[ { "version": "v1", "created": "Wed, 30 Jan 2013 15:02:44 GMT" } ]	2013-02-01T00:00:00	[ [ "Breese", "John S.", "" ], [ "Heckerman", "David", "" ], [ "Kadie", "Carl", "" ] ]	TITLE: Empirical Analysis of Predictive Algorithms for Collaborative Filtering ABSTRACT: Collaborative filtering or recommender systems use a database about user preferences to predict additional topics or products a new user might like. In this paper we describe several algorithms designed for this task, including techniques based on correlation coefficients, vector-based similarity calculations, and statistical Bayesian methods. We compare the predictive accuracy of the various methods in a set of representative problem domains. We use two basic classes of evaluation metrics. The first characterizes accuracy over a set of individual predictions in terms of average absolute deviation. The second estimates the utility of a ranked list of suggested items. This metric uses an estimate of the probability that a user will see a recommendation in an ordered list. Experiments were run for datasets associated with 3 application areas, 4 experimental protocols, and the 2 evaluation metrics for the various algorithms. Results indicate that for a wide range of conditions, Bayesian networks with decision trees at each node and correlation methods outperform Bayesian-clustering and vector-similarity methods. Between correlation and Bayesian networks, the preferred method depends on the nature of the dataset, nature of the application (ranked versus one-by-one presentation), and the availability of votes with which to make predictions. Other considerations include the size of database, speed of predictions, and learning time.
0903.4960	Michael Schreiber	Michael Schreiber	A Case Study of the Modified Hirsch Index hm Accounting for Multiple Co-authors	29 pages, including 2 tables, 3 figures with 7 plots altogether, accepted for publication in J. Am. Soc. Inf. Sci. Techn. vol. 60 (5) 2009	J. Am. Soc. Inf. Sci. Techn. 60, 1274-1282 (2009)	10.1002/asi.21057	null	physics.soc-ph	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	J. E. Hirsch (2005) introduced the h-index to quantify an individual's scientific research output by the largest number h of a scientist's papers, that received at least h citations. This so-called Hirsch index can be easily modified to take multiple co-authorship into account by counting the papers fractionally according to (the inverse of) the number of authors. I have worked out 26 empirical cases of physicists to illustrate the effect of this modification. Although the correlation between the original and the modified Hirsch index is relatively strong, the arrangement of the datasets is significantly different depending on whether they are put into order according to the values of either the original or the modified index.	[ { "version": "v1", "created": "Sat, 28 Mar 2009 10:01:42 GMT" } ]	2013-01-31T00:00:00	[ [ "Schreiber", "Michael", "" ] ]	TITLE: A Case Study of the Modified Hirsch Index hm Accounting for Multiple Co-authors ABSTRACT: J. E. Hirsch (2005) introduced the h-index to quantify an individual's scientific research output by the largest number h of a scientist's papers, that received at least h citations. This so-called Hirsch index can be easily modified to take multiple co-authorship into account by counting the papers fractionally according to (the inverse of) the number of authors. I have worked out 26 empirical cases of physicists to illustrate the effect of this modification. Although the correlation between the original and the modified Hirsch index is relatively strong, the arrangement of the datasets is significantly different depending on whether they are put into order according to the values of either the original or the modified index.
1202.3861	Michael Schreiber	Michael Schreiber	Inconsistencies of Recently Proposed Citation Impact Indicators and how to Avoid Them	14 pages, 9 figures, accepted by Journal of the American Society for Information Science and Technology Final version with slightly changed figures, new scoring rule, extended discussion	J. Am. Soc. Inf. Sci. Techn. 63(10), 2062-2073, (2012)	null	null	stat.AP cs.DL physics.soc-ph	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	It is shown that under certain circumstances in particular for small datasets the recently proposed citation impact indicators I3(6PR) and R(6,k) behave inconsistently when additional papers or citations are taken into consideration. Three simple examples are presented, in which the indicators fluctuate strongly and the ranking of scientists in the evaluated group is sometimes completely mixed up by minor changes in the data base. The erratic behavior is traced to the specific way in which weights are attributed to the six percentile rank classes, specifically for the tied papers. For 100 percentile rank classes the effects will be less serious. For the 6 classes it is demonstrated that a different way of assigning weights avoids these problems, although the non-linearity of the weights for the different percentile rank classes can still lead to (much less frequent) changes in the ranking. This behavior is not undesired, because it can be used to correct for differences in citation behavior in different fields. Remaining deviations from the theoretical value R(6,k) = 1.91 can be avoided by a new scoring rule, the fractional scoring. Previously proposed consistency criteria are amended by another property of strict independence which a performance indicator should aim at.	[ { "version": "v1", "created": "Fri, 17 Feb 2012 10:05:04 GMT" }, { "version": "v2", "created": "Wed, 4 Apr 2012 08:33:52 GMT" } ]	2013-01-31T00:00:00	[ [ "Schreiber", "Michael", "" ] ]	TITLE: Inconsistencies of Recently Proposed Citation Impact Indicators and how to Avoid Them ABSTRACT: It is shown that under certain circumstances in particular for small datasets the recently proposed citation impact indicators I3(6PR) and R(6,k) behave inconsistently when additional papers or citations are taken into consideration. Three simple examples are presented, in which the indicators fluctuate strongly and the ranking of scientists in the evaluated group is sometimes completely mixed up by minor changes in the data base. The erratic behavior is traced to the specific way in which weights are attributed to the six percentile rank classes, specifically for the tied papers. For 100 percentile rank classes the effects will be less serious. For the 6 classes it is demonstrated that a different way of assigning weights avoids these problems, although the non-linearity of the weights for the different percentile rank classes can still lead to (much less frequent) changes in the ranking. This behavior is not undesired, because it can be used to correct for differences in citation behavior in different fields. Remaining deviations from the theoretical value R(6,k) = 1.91 can be avoided by a new scoring rule, the fractional scoring. Previously proposed consistency criteria are amended by another property of strict independence which a performance indicator should aim at.
1202.4605	Andreas Raue	Andreas Raue, Clemens Kreutz, Fabian Joachim Theis, Jens Timmer	Joining Forces of Bayesian and Frequentist Methodology: A Study for Inference in the Presence of Non-Identifiability	Article to appear in Phil. Trans. Roy. Soc. A	Phil. Trans. R. Soc. A. 371, 20110544, 2013	10.1098/rsta.2011.0544	null	physics.data-an	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Increasingly complex applications involve large datasets in combination with non-linear and high dimensional mathematical models. In this context, statistical inference is a challenging issue that calls for pragmatic approaches that take advantage of both Bayesian and frequentist methods. The elegance of Bayesian methodology is founded in the propagation of information content provided by experimental data and prior assumptions to the posterior probability distribution of model predictions. However, for complex applications experimental data and prior assumptions potentially constrain the posterior probability distribution insufficiently. In these situations Bayesian Markov chain Monte Carlo sampling can be infeasible. From a frequentist point of view insufficient experimental data and prior assumptions can be interpreted as non-identifiability. The profile likelihood approach offers to detect and to resolve non-identifiability by experimental design iteratively. Therefore, it allows one to better constrain the posterior probability distribution until Markov chain Monte Carlo sampling can be used securely. Using an application from cell biology we compare both methods and show that a successive application of both methods facilitates a realistic assessment of uncertainty in model predictions.	[ { "version": "v1", "created": "Tue, 21 Feb 2012 11:44:06 GMT" } ]	2013-01-31T00:00:00	[ [ "Raue", "Andreas", "" ], [ "Kreutz", "Clemens", "" ], [ "Theis", "Fabian Joachim", "" ], [ "Timmer", "Jens", "" ] ]	TITLE: Joining Forces of Bayesian and Frequentist Methodology: A Study for Inference in the Presence of Non-Identifiability ABSTRACT: Increasingly complex applications involve large datasets in combination with non-linear and high dimensional mathematical models. In this context, statistical inference is a challenging issue that calls for pragmatic approaches that take advantage of both Bayesian and frequentist methods. The elegance of Bayesian methodology is founded in the propagation of information content provided by experimental data and prior assumptions to the posterior probability distribution of model predictions. However, for complex applications experimental data and prior assumptions potentially constrain the posterior probability distribution insufficiently. In these situations Bayesian Markov chain Monte Carlo sampling can be infeasible. From a frequentist point of view insufficient experimental data and prior assumptions can be interpreted as non-identifiability. The profile likelihood approach offers to detect and to resolve non-identifiability by experimental design iteratively. Therefore, it allows one to better constrain the posterior probability distribution until Markov chain Monte Carlo sampling can be used securely. Using an application from cell biology we compare both methods and show that a successive application of both methods facilitates a realistic assessment of uncertainty in model predictions.
1211.2756	Anton Korobeynikov	Sergey I. Nikolenko, Anton I. Korobeynikov and Max A. Alekseyev	BayesHammer: Bayesian clustering for error correction in single-cell sequencing	null	BMC Genomics 14(Suppl 1) (2013), pp. S7	10.1186/1471-2164-14-S1-S7	null	q-bio.QM cs.CE cs.DS q-bio.GN	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Error correction of sequenced reads remains a difficult task, especially in single-cell sequencing projects with extremely non-uniform coverage. While existing error correction tools designed for standard (multi-cell) sequencing data usually come up short in single-cell sequencing projects, algorithms actually used for single-cell error correction have been so far very simplistic. We introduce several novel algorithms based on Hamming graphs and Bayesian subclustering in our new error correction tool BayesHammer. While BayesHammer was designed for single-cell sequencing, we demonstrate that it also improves on existing error correction tools for multi-cell sequencing data while working much faster on real-life datasets. We benchmark BayesHammer on both $k$-mer counts and actual assembly results with the SPAdes genome assembler.	[ { "version": "v1", "created": "Mon, 12 Nov 2012 19:52:34 GMT" } ]	2013-01-31T00:00:00	[ [ "Nikolenko", "Sergey I.", "" ], [ "Korobeynikov", "Anton I.", "" ], [ "Alekseyev", "Max A.", "" ] ]	TITLE: BayesHammer: Bayesian clustering for error correction in single-cell sequencing ABSTRACT: Error correction of sequenced reads remains a difficult task, especially in single-cell sequencing projects with extremely non-uniform coverage. While existing error correction tools designed for standard (multi-cell) sequencing data usually come up short in single-cell sequencing projects, algorithms actually used for single-cell error correction have been so far very simplistic. We introduce several novel algorithms based on Hamming graphs and Bayesian subclustering in our new error correction tool BayesHammer. While BayesHammer was designed for single-cell sequencing, we demonstrate that it also improves on existing error correction tools for multi-cell sequencing data while working much faster on real-life datasets. We benchmark BayesHammer on both $k$-mer counts and actual assembly results with the SPAdes genome assembler.
0910.5260	Sewoong Oh	Raghunandan H. Keshavan, Sewoong Oh	A Gradient Descent Algorithm on the Grassman Manifold for Matrix Completion	26 pages, 15 figures	null	10.1016/j.trc.2012.12.007	null	cs.NA cs.LG	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	We consider the problem of reconstructing a low-rank matrix from a small subset of its entries. In this paper, we describe the implementation of an efficient algorithm called OptSpace, based on singular value decomposition followed by local manifold optimization, for solving the low-rank matrix completion problem. It has been shown that if the number of revealed entries is large enough, the output of singular value decomposition gives a good estimate for the original matrix, so that local optimization reconstructs the correct matrix with high probability. We present numerical results which show that this algorithm can reconstruct the low rank matrix exactly from a very small subset of its entries. We further study the robustness of the algorithm with respect to noise, and its performance on actual collaborative filtering datasets.	[ { "version": "v1", "created": "Tue, 27 Oct 2009 22:19:31 GMT" }, { "version": "v2", "created": "Tue, 3 Nov 2009 23:35:13 GMT" } ]	2013-01-30T00:00:00	[ [ "Keshavan", "Raghunandan H.", "" ], [ "Oh", "Sewoong", "" ] ]	TITLE: A Gradient Descent Algorithm on the Grassman Manifold for Matrix Completion ABSTRACT: We consider the problem of reconstructing a low-rank matrix from a small subset of its entries. In this paper, we describe the implementation of an efficient algorithm called OptSpace, based on singular value decomposition followed by local manifold optimization, for solving the low-rank matrix completion problem. It has been shown that if the number of revealed entries is large enough, the output of singular value decomposition gives a good estimate for the original matrix, so that local optimization reconstructs the correct matrix with high probability. We present numerical results which show that this algorithm can reconstruct the low rank matrix exactly from a very small subset of its entries. We further study the robustness of the algorithm with respect to noise, and its performance on actual collaborative filtering datasets.
1205.4377	Kirill Trapeznikov	Kirill Trapeznikov, Venkatesh Saligrama, David Castanon	Multi-Stage Classifier Design	null	null	null	null	cs.CV stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	In many classification systems, sensing modalities have different acquisition costs. It is often {\it unnecessary} to use every modality to classify a majority of examples. We study a multi-stage system in a prediction time cost reduction setting, where the full data is available for training, but for a test example, measurements in a new modality can be acquired at each stage for an additional cost. We seek decision rules to reduce the average measurement acquisition cost. We formulate an empirical risk minimization problem (ERM) for a multi-stage reject classifier, wherein the stage $k$ classifier either classifies a sample using only the measurements acquired so far or rejects it to the next stage where more attributes can be acquired for a cost. To solve the ERM problem, we show that the optimal reject classifier at each stage is a combination of two binary classifiers, one biased towards positive examples and the other biased towards negative examples. We use this parameterization to construct stage-by-stage global surrogate risk, develop an iterative algorithm in the boosting framework and present convergence and generalization results. We test our work on synthetic, medical and explosives detection datasets. Our results demonstrate that substantial cost reduction without a significant sacrifice in accuracy is achievable.	[ { "version": "v1", "created": "Sun, 20 May 2012 03:15:13 GMT" }, { "version": "v2", "created": "Tue, 29 Jan 2013 16:54:30 GMT" } ]	2013-01-30T00:00:00	[ [ "Trapeznikov", "Kirill", "" ], [ "Saligrama", "Venkatesh", "" ], [ "Castanon", "David", "" ] ]	TITLE: Multi-Stage Classifier Design ABSTRACT: In many classification systems, sensing modalities have different acquisition costs. It is often {\it unnecessary} to use every modality to classify a majority of examples. We study a multi-stage system in a prediction time cost reduction setting, where the full data is available for training, but for a test example, measurements in a new modality can be acquired at each stage for an additional cost. We seek decision rules to reduce the average measurement acquisition cost. We formulate an empirical risk minimization problem (ERM) for a multi-stage reject classifier, wherein the stage $k$ classifier either classifies a sample using only the measurements acquired so far or rejects it to the next stage where more attributes can be acquired for a cost. To solve the ERM problem, we show that the optimal reject classifier at each stage is a combination of two binary classifiers, one biased towards positive examples and the other biased towards negative examples. We use this parameterization to construct stage-by-stage global surrogate risk, develop an iterative algorithm in the boosting framework and present convergence and generalization results. We test our work on synthetic, medical and explosives detection datasets. Our results demonstrate that substantial cost reduction without a significant sacrifice in accuracy is achievable.
1301.6686	Gregory F. Cooper	Gregory F. Cooper, Changwon Yoo	Causal Discovery from a Mixture of Experimental and Observational Data	Appears in Proceedings of the Fifteenth Conference on Uncertainty in Artificial Intelligence (UAI1999)	null	null	UAI-P-1999-PG-116-125	cs.AI	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	This paper describes a Bayesian method for combining an arbitrary mixture of observational and experimental data in order to learn causal Bayesian networks. Observational data are passively observed. Experimental data, such as that produced by randomized controlled trials, result from the experimenter manipulating one or more variables (typically randomly) and observing the states of other variables. The paper presents a Bayesian method for learning the causal structure and parameters of the underlying causal process that is generating the data, given that (1) the data contains a mixture of observational and experimental case records, and (2) the causal process is modeled as a causal Bayesian network. This learning method was applied using as input various mixtures of experimental and observational data that were generated from the ALARM causal Bayesian network. In these experiments, the absolute and relative quantities of experimental and observational data were varied systematically. For each of these training datasets, the learning method was applied to predict the causal structure and to estimate the causal parameters that exist among randomly selected pairs of nodes in ALARM that are not confounded. The paper reports how these structure predictions and parameter estimates compare with the true causal structures and parameters as given by the ALARM network.	[ { "version": "v1", "created": "Wed, 23 Jan 2013 15:57:22 GMT" } ]	2013-01-30T00:00:00	[ [ "Cooper", "Gregory F.", "" ], [ "Yoo", "Changwon", "" ] ]	TITLE: Causal Discovery from a Mixture of Experimental and Observational Data ABSTRACT: This paper describes a Bayesian method for combining an arbitrary mixture of observational and experimental data in order to learn causal Bayesian networks. Observational data are passively observed. Experimental data, such as that produced by randomized controlled trials, result from the experimenter manipulating one or more variables (typically randomly) and observing the states of other variables. The paper presents a Bayesian method for learning the causal structure and parameters of the underlying causal process that is generating the data, given that (1) the data contains a mixture of observational and experimental case records, and (2) the causal process is modeled as a causal Bayesian network. This learning method was applied using as input various mixtures of experimental and observational data that were generated from the ALARM causal Bayesian network. In these experiments, the absolute and relative quantities of experimental and observational data were varied systematically. For each of these training datasets, the learning method was applied to predict the causal structure and to estimate the causal parameters that exist among randomly selected pairs of nodes in ALARM that are not confounded. The paper reports how these structure predictions and parameter estimates compare with the true causal structures and parameters as given by the ALARM network.
1301.6723	Stefano Monti	Stefano Monti, Gregory F. Cooper	A Bayesian Network Classifier that Combines a Finite Mixture Model and a Naive Bayes Model	Appears in Proceedings of the Fifteenth Conference on Uncertainty in Artificial Intelligence (UAI1999)	null	null	UAI-P-1999-PG-447-456	cs.LG cs.AI stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	In this paper we present a new Bayesian network model for classification that combines the naive-Bayes (NB) classifier and the finite-mixture (FM) classifier. The resulting classifier aims at relaxing the strong assumptions on which the two component models are based, in an attempt to improve on their classification performance, both in terms of accuracy and in terms of calibration of the estimated probabilities. The proposed classifier is obtained by superimposing a finite mixture model on the set of feature variables of a naive Bayes model. We present experimental results that compare the predictive performance on real datasets of the new classifier with the predictive performance of the NB classifier and the FM classifier.	[ { "version": "v1", "created": "Wed, 23 Jan 2013 15:59:54 GMT" } ]	2013-01-30T00:00:00	[ [ "Monti", "Stefano", "" ], [ "Cooper", "Gregory F.", "" ] ]	TITLE: A Bayesian Network Classifier that Combines a Finite Mixture Model and a Naive Bayes Model ABSTRACT: In this paper we present a new Bayesian network model for classification that combines the naive-Bayes (NB) classifier and the finite-mixture (FM) classifier. The resulting classifier aims at relaxing the strong assumptions on which the two component models are based, in an attempt to improve on their classification performance, both in terms of accuracy and in terms of calibration of the estimated probabilities. The proposed classifier is obtained by superimposing a finite mixture model on the set of feature variables of a naive Bayes model. We present experimental results that compare the predictive performance on real datasets of the new classifier with the predictive performance of the NB classifier and the FM classifier.
1301.6770	Zhixiang Eddie Xu	Zhixiang (Eddie) Xu, Minmin Chen, Kilian Q. Weinberger, Fei Sha	An alternative text representation to TF-IDF and Bag-of-Words	null	null	null	null	cs.IR cs.LG stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	In text mining, information retrieval, and machine learning, text documents are commonly represented through variants of sparse Bag of Words (sBoW) vectors (e.g. TF-IDF). Although simple and intuitive, sBoW style representations suffer from their inherent over-sparsity and fail to capture word-level synonymy and polysemy. Especially when labeled data is limited (e.g. in document classification), or the text documents are short (e.g. emails or abstracts), many features are rarely observed within the training corpus. This leads to overfitting and reduced generalization accuracy. In this paper we propose Dense Cohort of Terms (dCoT), an unsupervised algorithm to learn improved sBoW document features. dCoT explicitly models absent words by removing and reconstructing random sub-sets of words in the unlabeled corpus. With this approach, dCoT learns to reconstruct frequent words from co-occurring infrequent words and maps the high dimensional sparse sBoW vectors into a low-dimensional dense representation. We show that the feature removal can be marginalized out and that the reconstruction can be solved for in closed-form. We demonstrate empirically, on several benchmark datasets, that dCoT features significantly improve the classification accuracy across several document classification tasks.	[ { "version": "v1", "created": "Mon, 28 Jan 2013 21:04:45 GMT" } ]	2013-01-30T00:00:00	[ [ "Zhixiang", "", "", "Eddie" ], [ "Xu", "", "" ], [ "Chen", "Minmin", "" ], [ "Weinberger", "Kilian Q.", "" ], [ "Sha", "Fei", "" ] ]	TITLE: An alternative text representation to TF-IDF and Bag-of-Words ABSTRACT: In text mining, information retrieval, and machine learning, text documents are commonly represented through variants of sparse Bag of Words (sBoW) vectors (e.g. TF-IDF). Although simple and intuitive, sBoW style representations suffer from their inherent over-sparsity and fail to capture word-level synonymy and polysemy. Especially when labeled data is limited (e.g. in document classification), or the text documents are short (e.g. emails or abstracts), many features are rarely observed within the training corpus. This leads to overfitting and reduced generalization accuracy. In this paper we propose Dense Cohort of Terms (dCoT), an unsupervised algorithm to learn improved sBoW document features. dCoT explicitly models absent words by removing and reconstructing random sub-sets of words in the unlabeled corpus. With this approach, dCoT learns to reconstruct frequent words from co-occurring infrequent words and maps the high dimensional sparse sBoW vectors into a low-dimensional dense representation. We show that the feature removal can be marginalized out and that the reconstruction can be solved for in closed-form. We demonstrate empirically, on several benchmark datasets, that dCoT features significantly improve the classification accuracy across several document classification tasks.
1301.6800	Mokhov, Nikolai	C. Yoshikawa (Muons, Inc.), A. Leveling, N.V. Mokhov, J. Morgan, D. Neuffer, S. Striganov (Fermilab)	Optimization of the Target Subsystem for the New g-2 Experiment	4 pp. 3rd International Particle Accelerator Conference (IPAC 2012) 20-25 May 2012, New Orleans, Louisiana	null	null	FERMILAB-CONF-12-202-AD-APC	physics.acc-ph	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	A precision measurement of the muon anomalous magnetic moment, $a_{\mu} = (g-2)/2$, was previously performed at BNL with a result of 2.2 - 2.7 standard deviations above the Standard Model (SM) theoretical calculations. The same experimental apparatus is being planned to run in the new Muon Campus at Fermilab, where the muon beam is expected to have less pion contamination and the extended dataset may provide a possible $7.5\sigma$ deviation from the SM, creating a sensitive and complementary bench mark for proposed SM extensions. We report here on a preliminary study of the target subsystem where the apparatus is optimized for pions that have favorable phase space to create polarized daughter muons around the magic momentum of 3.094 GeV/c, which is needed by the downstream g 2 muon ring.	[ { "version": "v1", "created": "Mon, 28 Jan 2013 22:30:19 GMT" } ]	2013-01-30T00:00:00	[ [ "Yoshikawa", "C.", "", "Muons, Inc." ], [ "Leveling", "A.", "", "Fermilab" ], [ "Mokhov", "N. V.", "", "Fermilab" ], [ "Morgan", "J.", "", "Fermilab" ], [ "Neuffer", "D.", "", "Fermilab" ], [ "Striganov", "S.", "", "Fermilab" ] ]	TITLE: Optimization of the Target Subsystem for the New g-2 Experiment ABSTRACT: A precision measurement of the muon anomalous magnetic moment, $a_{\mu} = (g-2)/2$, was previously performed at BNL with a result of 2.2 - 2.7 standard deviations above the Standard Model (SM) theoretical calculations. The same experimental apparatus is being planned to run in the new Muon Campus at Fermilab, where the muon beam is expected to have less pion contamination and the extended dataset may provide a possible $7.5\sigma$ deviation from the SM, creating a sensitive and complementary bench mark for proposed SM extensions. We report here on a preliminary study of the target subsystem where the apparatus is optimized for pions that have favorable phase space to create polarized daughter muons around the magic momentum of 3.094 GeV/c, which is needed by the downstream g 2 muon ring.
1301.6870	Paridhi Jain	Anshu Malhotra, Luam Totti, Wagner Meira Jr., Ponnurangam Kumaraguru, Virgilio Almeida	Studying User Footprints in Different Online Social Networks	The paper is already published in ASONAM 2012	null	null	null	cs.SI	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	With the growing popularity and usage of online social media services, people now have accounts (some times several) on multiple and diverse services like Facebook, LinkedIn, Twitter and YouTube. Publicly available information can be used to create a digital footprint of any user using these social media services. Generating such digital footprints can be very useful for personalization, profile management, detecting malicious behavior of users. A very important application of analyzing users' online digital footprints is to protect users from potential privacy and security risks arising from the huge publicly available user information. We extracted information about user identities on different social networks through Social Graph API, FriendFeed, and Profilactic; we collated our own dataset to create the digital footprints of the users. We used username, display name, description, location, profile image, and number of connections to generate the digital footprints of the user. We applied context specific techniques (e.g. Jaro Winkler similarity, Wordnet based ontologies) to measure the similarity of the user profiles on different social networks. We specifically focused on Twitter and LinkedIn. In this paper, we present the analysis and results from applying automated classifiers for disambiguating profiles belonging to the same user from different social networks. UserID and Name were found to be the most discriminative features for disambiguating user profiles. Using the most promising set of features and similarity metrics, we achieved accuracy, precision and recall of 98%, 99%, and 96%, respectively.	[ { "version": "v1", "created": "Tue, 29 Jan 2013 09:29:54 GMT" } ]	2013-01-30T00:00:00	[ [ "Malhotra", "Anshu", "" ], [ "Totti", "Luam", "" ], [ "Meira", "Wagner", "Jr." ], [ "Kumaraguru", "Ponnurangam", "" ], [ "Almeida", "Virgilio", "" ] ]	TITLE: Studying User Footprints in Different Online Social Networks ABSTRACT: With the growing popularity and usage of online social media services, people now have accounts (some times several) on multiple and diverse services like Facebook, LinkedIn, Twitter and YouTube. Publicly available information can be used to create a digital footprint of any user using these social media services. Generating such digital footprints can be very useful for personalization, profile management, detecting malicious behavior of users. A very important application of analyzing users' online digital footprints is to protect users from potential privacy and security risks arising from the huge publicly available user information. We extracted information about user identities on different social networks through Social Graph API, FriendFeed, and Profilactic; we collated our own dataset to create the digital footprints of the users. We used username, display name, description, location, profile image, and number of connections to generate the digital footprints of the user. We applied context specific techniques (e.g. Jaro Winkler similarity, Wordnet based ontologies) to measure the similarity of the user profiles on different social networks. We specifically focused on Twitter and LinkedIn. In this paper, we present the analysis and results from applying automated classifiers for disambiguating profiles belonging to the same user from different social networks. UserID and Name were found to be the most discriminative features for disambiguating user profiles. Using the most promising set of features and similarity metrics, we achieved accuracy, precision and recall of 98%, 99%, and 96%, respectively.
1210.0137	Pierre Deville Pierre	Vincent D. Blondel, Markus Esch, Connie Chan, Fabrice Clerot, Pierre Deville, Etienne Huens, Fr\'ed\'eric Morlot, Zbigniew Smoreda and Cezary Ziemlicki	Data for Development: the D4D Challenge on Mobile Phone Data	10 pages, 3 figures	null	null	null	cs.CY cs.SI physics.soc-ph stat.CO	http://creativecommons.org/licenses/by-nc-sa/3.0/	The Orange "Data for Development" (D4D) challenge is an open data challenge on anonymous call patterns of Orange's mobile phone users in Ivory Coast. The goal of the challenge is to help address society development questions in novel ways by contributing to the socio-economic development and well-being of the Ivory Coast population. Participants to the challenge are given access to four mobile phone datasets and the purpose of this paper is to describe the four datasets. The website http://www.d4d.orange.com contains more information about the participation rules. The datasets are based on anonymized Call Detail Records (CDR) of phone calls and SMS exchanges between five million of Orange's customers in Ivory Coast between December 1, 2011 and April 28, 2012. The datasets are: (a) antenna-to-antenna traffic on an hourly basis, (b) individual trajectories for 50,000 customers for two week time windows with antenna location information, (3) individual trajectories for 500,000 customers over the entire observation period with sub-prefecture location information, and (4) a sample of communication graphs for 5,000 customers	[ { "version": "v1", "created": "Sat, 29 Sep 2012 17:39:16 GMT" }, { "version": "v2", "created": "Mon, 28 Jan 2013 12:56:55 GMT" } ]	2013-01-29T00:00:00	[ [ "Blondel", "Vincent D.", "" ], [ "Esch", "Markus", "" ], [ "Chan", "Connie", "" ], [ "Clerot", "Fabrice", "" ], [ "Deville", "Pierre", "" ], [ "Huens", "Etienne", "" ], [ "Morlot", "Frédéric", "" ], [ "Smoreda", "Zbigniew", "" ], [ "Ziemlicki", "Cezary", "" ] ]	TITLE: Data for Development: the D4D Challenge on Mobile Phone Data ABSTRACT: The Orange "Data for Development" (D4D) challenge is an open data challenge on anonymous call patterns of Orange's mobile phone users in Ivory Coast. The goal of the challenge is to help address society development questions in novel ways by contributing to the socio-economic development and well-being of the Ivory Coast population. Participants to the challenge are given access to four mobile phone datasets and the purpose of this paper is to describe the four datasets. The website http://www.d4d.orange.com contains more information about the participation rules. The datasets are based on anonymized Call Detail Records (CDR) of phone calls and SMS exchanges between five million of Orange's customers in Ivory Coast between December 1, 2011 and April 28, 2012. The datasets are: (a) antenna-to-antenna traffic on an hourly basis, (b) individual trajectories for 50,000 customers for two week time windows with antenna location information, (3) individual trajectories for 500,000 customers over the entire observation period with sub-prefecture location information, and (4) a sample of communication graphs for 5,000 customers
1301.4293	Limin Yao	Sebastian Riedel, Limin Yao, Andrew McCallum	Latent Relation Representations for Universal Schemas	4 pages, ICLR workshop	null	null	null	cs.LG stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Traditional relation extraction predicts relations within some fixed and finite target schema. Machine learning approaches to this task require either manual annotation or, in the case of distant supervision, existing structured sources of the same schema. The need for existing datasets can be avoided by using a universal schema: the union of all involved schemas (surface form predicates as in OpenIE, and relations in the schemas of pre-existing databases). This schema has an almost unlimited set of relations (due to surface forms), and supports integration with existing structured data (through the relation types of existing databases). To populate a database of such schema we present a family of matrix factorization models that predict affinity between database tuples and relations. We show that this achieves substantially higher accuracy than the traditional classification approach. More importantly, by operating simultaneously on relations observed in text and in pre-existing structured DBs such as Freebase, we are able to reason about unstructured and structured data in mutually-supporting ways. By doing so our approach outperforms state-of-the-art distant supervision systems.	[ { "version": "v1", "created": "Fri, 18 Jan 2013 04:37:30 GMT" }, { "version": "v2", "created": "Mon, 28 Jan 2013 20:10:21 GMT" } ]	2013-01-29T00:00:00	[ [ "Riedel", "Sebastian", "" ], [ "Yao", "Limin", "" ], [ "McCallum", "Andrew", "" ] ]	TITLE: Latent Relation Representations for Universal Schemas ABSTRACT: Traditional relation extraction predicts relations within some fixed and finite target schema. Machine learning approaches to this task require either manual annotation or, in the case of distant supervision, existing structured sources of the same schema. The need for existing datasets can be avoided by using a universal schema: the union of all involved schemas (surface form predicates as in OpenIE, and relations in the schemas of pre-existing databases). This schema has an almost unlimited set of relations (due to surface forms), and supports integration with existing structured data (through the relation types of existing databases). To populate a database of such schema we present a family of matrix factorization models that predict affinity between database tuples and relations. We show that this achieves substantially higher accuracy than the traditional classification approach. More importantly, by operating simultaneously on relations observed in text and in pre-existing structured DBs such as Freebase, we are able to reason about unstructured and structured data in mutually-supporting ways. By doing so our approach outperforms state-of-the-art distant supervision systems.
1301.5686	Jeon-Hyung Kang	Jeon-Hyung Kang, Jun Ma, Yan Liu	Transfer Topic Modeling with Ease and Scalability	2012 SIAM International Conference on Data Mining (SDM12) Pages: {564-575}	null	null	null	cs.CL cs.LG stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	The increasing volume of short texts generated on social media sites, such as Twitter or Facebook, creates a great demand for effective and efficient topic modeling approaches. While latent Dirichlet allocation (LDA) can be applied, it is not optimal due to its weakness in handling short texts with fast-changing topics and scalability concerns. In this paper, we propose a transfer learning approach that utilizes abundant labeled documents from other domains (such as Yahoo! News or Wikipedia) to improve topic modeling, with better model fitting and result interpretation. Specifically, we develop Transfer Hierarchical LDA (thLDA) model, which incorporates the label information from other domains via informative priors. In addition, we develop a parallel implementation of our model for large-scale applications. We demonstrate the effectiveness of our thLDA model on both a microblogging dataset and standard text collections including AP and RCV1 datasets.	[ { "version": "v1", "created": "Thu, 24 Jan 2013 02:02:13 GMT" }, { "version": "v2", "created": "Sat, 26 Jan 2013 18:00:19 GMT" } ]	2013-01-29T00:00:00	[ [ "Kang", "Jeon-Hyung", "" ], [ "Ma", "Jun", "" ], [ "Liu", "Yan", "" ] ]	TITLE: Transfer Topic Modeling with Ease and Scalability ABSTRACT: The increasing volume of short texts generated on social media sites, such as Twitter or Facebook, creates a great demand for effective and efficient topic modeling approaches. While latent Dirichlet allocation (LDA) can be applied, it is not optimal due to its weakness in handling short texts with fast-changing topics and scalability concerns. In this paper, we propose a transfer learning approach that utilizes abundant labeled documents from other domains (such as Yahoo! News or Wikipedia) to improve topic modeling, with better model fitting and result interpretation. Specifically, we develop Transfer Hierarchical LDA (thLDA) model, which incorporates the label information from other domains via informative priors. In addition, we develop a parallel implementation of our model for large-scale applications. We demonstrate the effectiveness of our thLDA model on both a microblogging dataset and standard text collections including AP and RCV1 datasets.
1301.6553	Thomas Couronne	Thomas Couronne, Zbigniew Smoreda, Ana-Maria Olteanu	Chatty Mobiles:Individual mobility and communication patterns	NetMob 2011, Boston	null	null	null	cs.CY	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Human mobility analysis is an important issue in social sciences, and mobility data are among the most sought-after sources of information in ur- Data ban studies, geography, transportation and territory management. In network sciences mobility studies have become popular in the past few years, especially using mobile phone location data. For preserving the customer privacy, datasets furnished by telecom operators are anonymized. At the same time, the large size of datasets often makes the task of calculating all observed trajectories very difficult and time-consuming. One solution could be to sample users. However, the fact of not having information about the mobile user makes the sampling delicate. Some researchers select randomly a sample of users from their dataset. Others try to optimize this method, for example, taking into account only users with a certain number or frequency of locations recorded. At the first glance, the second choice seems to be more efficient: having more individual traces makes the analysis more precise. However, the most frequently used CDR data (Call Detail Records) have location generated only at the moment of communication (call, SMS, data connection). Due to this fact, users mobility patterns cannot be precisely built upon their communication patterns. Hence, these data have evident short-comings both in terms of spatial and temporal scale. In this paper we propose to estimate the correlation between the users communication and mo- bility in order to better assess the bias of frequency based sampling. Using technical GSM network data (including communication but also independent mobility records), we will analyze the relationship between communication and mobility patterns.	[ { "version": "v1", "created": "Mon, 28 Jan 2013 14:19:48 GMT" } ]	2013-01-29T00:00:00	[ [ "Couronne", "Thomas", "" ], [ "Smoreda", "Zbigniew", "" ], [ "Olteanu", "Ana-Maria", "" ] ]	TITLE: Chatty Mobiles:Individual mobility and communication patterns ABSTRACT: Human mobility analysis is an important issue in social sciences, and mobility data are among the most sought-after sources of information in ur- Data ban studies, geography, transportation and territory management. In network sciences mobility studies have become popular in the past few years, especially using mobile phone location data. For preserving the customer privacy, datasets furnished by telecom operators are anonymized. At the same time, the large size of datasets often makes the task of calculating all observed trajectories very difficult and time-consuming. One solution could be to sample users. However, the fact of not having information about the mobile user makes the sampling delicate. Some researchers select randomly a sample of users from their dataset. Others try to optimize this method, for example, taking into account only users with a certain number or frequency of locations recorded. At the first glance, the second choice seems to be more efficient: having more individual traces makes the analysis more precise. However, the most frequently used CDR data (Call Detail Records) have location generated only at the moment of communication (call, SMS, data connection). Due to this fact, users mobility patterns cannot be precisely built upon their communication patterns. Hence, these data have evident short-comings both in terms of spatial and temporal scale. In this paper we propose to estimate the correlation between the users communication and mo- bility in order to better assess the bias of frequency based sampling. Using technical GSM network data (including communication but also independent mobility records), we will analyze the relationship between communication and mobility patterns.
1301.5943	Lu\'is Filipe Te\'ofilo	Lu\'is Filipe Te\'ofilo, Luis Paulo Reis	Identifying Player\'s Strategies in No Limit Texas Hold\'em Poker through the Analysis of Individual Moves	null	null	null	null	cs.AI cs.GT	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	The development of competitive artificial Poker playing agents has proven to be a challenge, because agents must deal with unreliable information and deception which make it essential to model the opponents in order to achieve good results. This paper presents a methodology to develop opponent modeling techniques for Poker agents. The approach is based on applying clustering algorithms to a Poker game database in order to identify player types based on their actions. First, common game moves were identified by clustering all players\' moves. Then, player types were defined by calculating the frequency with which the players perform each type of movement. With the given dataset, 7 different types of players were identified with each one having at least one tactic that characterizes him. The identification of player types may improve the overall performance of Poker agents, because it helps the agents to predict the opponent\'s moves, by associating each opponent to a distinct cluster.	[ { "version": "v1", "created": "Fri, 25 Jan 2013 01:49:15 GMT" } ]	2013-01-28T00:00:00	[ [ "Teófilo", "Luís Filipe", "" ], [ "Reis", "Luis Paulo", "" ] ]	TITLE: Identifying Player\'s Strategies in No Limit Texas Hold\'em Poker through the Analysis of Individual Moves ABSTRACT: The development of competitive artificial Poker playing agents has proven to be a challenge, because agents must deal with unreliable information and deception which make it essential to model the opponents in order to achieve good results. This paper presents a methodology to develop opponent modeling techniques for Poker agents. The approach is based on applying clustering algorithms to a Poker game database in order to identify player types based on their actions. First, common game moves were identified by clustering all players\' moves. Then, player types were defined by calculating the frequency with which the players perform each type of movement. With the given dataset, 7 different types of players were identified with each one having at least one tactic that characterizes him. The identification of player types may improve the overall performance of Poker agents, because it helps the agents to predict the opponent\'s moves, by associating each opponent to a distinct cluster.
1212.2142	Arnab Chatterjee	Arnab Chatterjee, Marija Mitrovi\'c and Santo Fortunato	Universality in voting behavior: an empirical analysis	19 pages, 10 figures, 8 tables. The elections data-sets can be downloaded from http://becs.aalto.fi/en/research/complex_systems/elections/	Scientific Reports 3, 1049 (2013)	null	null	physics.soc-ph cs.SI physics.data-an	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Election data represent a precious source of information to study human behavior at a large scale. In proportional elections with open lists, the number of votes received by a candidate, rescaled by the average performance of all competitors in the same party list, has the same distribution regardless of the country and the year of the election. Here we provide the first thorough assessment of this claim. We analyzed election datasets of 15 countries with proportional systems. We confirm that a class of nations with similar election rules fulfill the universality claim. Discrepancies from this trend in other countries with open-lists elections are always associated with peculiar differences in the election rules, which matter more than differences between countries and historical periods. Our analysis shows that the role of parties in the electoral performance of candidates is crucial: alternative scalings not taking into account party affiliations lead to poor results.	[ { "version": "v1", "created": "Mon, 10 Dec 2012 17:26:06 GMT" }, { "version": "v2", "created": "Thu, 24 Jan 2013 12:41:20 GMT" } ]	2013-01-25T00:00:00	[ [ "Chatterjee", "Arnab", "" ], [ "Mitrović", "Marija", "" ], [ "Fortunato", "Santo", "" ] ]	TITLE: Universality in voting behavior: an empirical analysis ABSTRACT: Election data represent a precious source of information to study human behavior at a large scale. In proportional elections with open lists, the number of votes received by a candidate, rescaled by the average performance of all competitors in the same party list, has the same distribution regardless of the country and the year of the election. Here we provide the first thorough assessment of this claim. We analyzed election datasets of 15 countries with proportional systems. We confirm that a class of nations with similar election rules fulfill the universality claim. Discrepancies from this trend in other countries with open-lists elections are always associated with peculiar differences in the election rules, which matter more than differences between countries and historical periods. Our analysis shows that the role of parties in the electoral performance of candidates is crucial: alternative scalings not taking into account party affiliations lead to poor results.
1301.5399	Hoda Sadat Ayatollahi Tabatabaii	Hoda S. Ayatollahi Tabatabaii, Hamid R. Rabiee, Mohammad Hossein Rohban, Mostafa Salehi	Incorporating Betweenness Centrality in Compressive Sensing for Congestion Detection	null	null	null	null	cs.NI	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	This paper presents a new Compressive Sensing (CS) scheme for detecting network congested links. We focus on decreasing the required number of measurements to detect all congested links in the context of network tomography. We have expanded the LASSO objective function by adding a new term corresponding to the prior knowledge based on the relationship between the congested links and the corresponding link Betweenness Centrality (BC). The accuracy of the proposed model is verified by simulations on two real datasets. The results demonstrate that our model outperformed the state-of-the-art CS based method with significant improvements in terms of F-Score.	[ { "version": "v1", "created": "Wed, 23 Jan 2013 04:12:08 GMT" } ]	2013-01-24T00:00:00	[ [ "Tabatabaii", "Hoda S. Ayatollahi", "" ], [ "Rabiee", "Hamid R.", "" ], [ "Rohban", "Mohammad Hossein", "" ], [ "Salehi", "Mostafa", "" ] ]	TITLE: Incorporating Betweenness Centrality in Compressive Sensing for Congestion Detection ABSTRACT: This paper presents a new Compressive Sensing (CS) scheme for detecting network congested links. We focus on decreasing the required number of measurements to detect all congested links in the context of network tomography. We have expanded the LASSO objective function by adding a new term corresponding to the prior knowledge based on the relationship between the congested links and the corresponding link Betweenness Centrality (BC). The accuracy of the proposed model is verified by simulations on two real datasets. The results demonstrate that our model outperformed the state-of-the-art CS based method with significant improvements in terms of F-Score.
1204.4491	Huy Nguyen	Huy Nguyen, Rong Zheng	On Budgeted Influence Maximization in Social Networks	Submitted to JSAC NS	null	null	null	cs.SI physics.soc-ph	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Given a budget and arbitrary cost for selecting each node, the budgeted influence maximization (BIM) problem concerns selecting a set of seed nodes to disseminate some information that maximizes the total number of nodes influenced (termed as influence spread) in social networks at a total cost no more than the budget. Our proposed seed selection algorithm for the BIM problem guarantees an approximation ratio of (1 - 1/sqrt(e)). The seed selection algorithm needs to calculate the influence spread of candidate seed sets, which is known to be #P-complex. Identifying the linkage between the computation of marginal probabilities in Bayesian networks and the influence spread, we devise efficient heuristic algorithms for the latter problem. Experiments using both large-scale social networks and synthetically generated networks demonstrate superior performance of the proposed algorithm with moderate computation costs. Moreover, synthetic datasets allow us to vary the network parameters and gain important insights on the impact of graph structures on the performance of different algorithms.	[ { "version": "v1", "created": "Thu, 19 Apr 2012 22:50:48 GMT" }, { "version": "v2", "created": "Thu, 2 Aug 2012 05:02:19 GMT" }, { "version": "v3", "created": "Tue, 22 Jan 2013 07:01:49 GMT" } ]	2013-01-23T00:00:00	[ [ "Nguyen", "Huy", "" ], [ "Zheng", "Rong", "" ] ]	TITLE: On Budgeted Influence Maximization in Social Networks ABSTRACT: Given a budget and arbitrary cost for selecting each node, the budgeted influence maximization (BIM) problem concerns selecting a set of seed nodes to disseminate some information that maximizes the total number of nodes influenced (termed as influence spread) in social networks at a total cost no more than the budget. Our proposed seed selection algorithm for the BIM problem guarantees an approximation ratio of (1 - 1/sqrt(e)). The seed selection algorithm needs to calculate the influence spread of candidate seed sets, which is known to be #P-complex. Identifying the linkage between the computation of marginal probabilities in Bayesian networks and the influence spread, we devise efficient heuristic algorithms for the latter problem. Experiments using both large-scale social networks and synthetically generated networks demonstrate superior performance of the proposed algorithm with moderate computation costs. Moreover, synthetic datasets allow us to vary the network parameters and gain important insights on the impact of graph structures on the performance of different algorithms.
1301.5088	Ian Goodfellow	Ian J. Goodfellow	Piecewise Linear Multilayer Perceptrons and Dropout	null	null	null	null	stat.ML cs.LG	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	We propose a new type of hidden layer for a multilayer perceptron, and demonstrate that it obtains the best reported performance for an MLP on the MNIST dataset.	[ { "version": "v1", "created": "Tue, 22 Jan 2013 07:10:34 GMT" } ]	2013-01-23T00:00:00	[ [ "Goodfellow", "Ian J.", "" ] ]	TITLE: Piecewise Linear Multilayer Perceptrons and Dropout ABSTRACT: We propose a new type of hidden layer for a multilayer perceptron, and demonstrate that it obtains the best reported performance for an MLP on the MNIST dataset.
1301.5121	Alex Averbuch	Alex Averbuch, Martin Neumann	Partitioning Graph Databases - A Quantitative Evaluation	null	null	null	null	cs.DB cs.DC	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Electronic data is growing at increasing rates, in both size and connectivity: the increasing presence of, and interest in, relationships between data. An example is the Twitter social network graph. Due to this growth demand is increasing for technologies that can process such data. Currently relational databases are the predominant technology, but they are poorly suited to processing connected data as they are optimized for index-intensive operations. Conversely, graph databases are optimized for graph computation. They link records by direct references, avoiding index lookups, and enabling retrieval of adjacent elements in constant time, regardless of graph size. However, as data volume increases these databases outgrow the resources of one computer and data partitioning becomes necessary. We evaluate the viability of using graph partitioning algorithms to partition graph databases. A prototype partitioned database was developed. Three partitioning algorithms explored and one implemented. Three graph datasets were used: two real and one synthetically generated. These were partitioned in various ways and the impact on database performance measured. We defined one synthetic access pattern per dataset and executed each on the partitioned datasets. Evaluation took place in a simulation environment, ensuring repeatability and allowing measurement of metrics like network traffic and load balance. Results show that compared to random partitioning the partitioning algorithm reduced traffic by 40-90%. Executing the algorithm intermittently during usage maintained partition quality, while requiring only 1% the computation of initial partitioning. Strong correlations were found between theoretic quality metrics and generated network traffic under non-uniform access patterns.	[ { "version": "v1", "created": "Tue, 22 Jan 2013 09:48:34 GMT" } ]	2013-01-23T00:00:00	[ [ "Averbuch", "Alex", "" ], [ "Neumann", "Martin", "" ] ]	TITLE: Partitioning Graph Databases - A Quantitative Evaluation ABSTRACT: Electronic data is growing at increasing rates, in both size and connectivity: the increasing presence of, and interest in, relationships between data. An example is the Twitter social network graph. Due to this growth demand is increasing for technologies that can process such data. Currently relational databases are the predominant technology, but they are poorly suited to processing connected data as they are optimized for index-intensive operations. Conversely, graph databases are optimized for graph computation. They link records by direct references, avoiding index lookups, and enabling retrieval of adjacent elements in constant time, regardless of graph size. However, as data volume increases these databases outgrow the resources of one computer and data partitioning becomes necessary. We evaluate the viability of using graph partitioning algorithms to partition graph databases. A prototype partitioned database was developed. Three partitioning algorithms explored and one implemented. Three graph datasets were used: two real and one synthetically generated. These were partitioned in various ways and the impact on database performance measured. We defined one synthetic access pattern per dataset and executed each on the partitioned datasets. Evaluation took place in a simulation environment, ensuring repeatability and allowing measurement of metrics like network traffic and load balance. Results show that compared to random partitioning the partitioning algorithm reduced traffic by 40-90%. Executing the algorithm intermittently during usage maintained partition quality, while requiring only 1% the computation of initial partitioning. Strong correlations were found between theoretic quality metrics and generated network traffic under non-uniform access patterns.
1012.4506	Taha Sochi	Taha Sochi	High Throughput Software for Powder Diffraction and its Application to Heterogeneous Catalysis	thesis, 202 pages, 95 figures, 6 tables	null	null	null	physics.data-an hep-ex physics.chem-ph physics.comp-ph physics.ins-det	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	In this thesis we investigate high throughput computational methods for processing large quantities of data collected from synchrotrons and their application to spectral analysis of powder diffraction data. We also present the main product of this PhD programme, specifically a software called 'EasyDD' developed by the author. This software was created to meet the increasing demand on data processing and analysis capabilities as required by modern detectors which produce huge quantities of data. Modern detectors coupled with the high intensity X-ray sources available at synchrotrons have led to the situation where datasets can be collected in ever shorter time scales and in ever larger numbers. Such large volumes of datasets pose a data processing bottleneck which augments with current and future instrument development. EasyDD has achieved its objectives and made significant contributions to scientific research. It can also be used as a model for more mature attempts in the future. EasyDD is currently in use by a number of researchers in a number of academic and research institutions to process high-energy diffraction data. These include data collected by different techniques such as Energy Dispersive Diffraction, Angle Dispersive Diffraction and Computer Aided Tomography. EasyDD has already been used in a number of published studies, and is currently in use by the High Energy X-Ray Imaging Technology project. The software was also used by the author to process and analyse datasets collected from synchrotron radiation facilities. In this regard, the thesis presents novel scientific research involving the use of EasyDD to handle large diffraction datasets in the study of alumina-supported metal oxide catalyst bodies. These data were collected using Tomographic Energy Dispersive Diffraction Imaging and Computer Aided Tomography techniques.	[ { "version": "v1", "created": "Mon, 20 Dec 2010 23:35:54 GMT" } ]	2013-01-22T00:00:00	[ [ "Sochi", "Taha", "" ] ]	TITLE: High Throughput Software for Powder Diffraction and its Application to Heterogeneous Catalysis ABSTRACT: In this thesis we investigate high throughput computational methods for processing large quantities of data collected from synchrotrons and their application to spectral analysis of powder diffraction data. We also present the main product of this PhD programme, specifically a software called 'EasyDD' developed by the author. This software was created to meet the increasing demand on data processing and analysis capabilities as required by modern detectors which produce huge quantities of data. Modern detectors coupled with the high intensity X-ray sources available at synchrotrons have led to the situation where datasets can be collected in ever shorter time scales and in ever larger numbers. Such large volumes of datasets pose a data processing bottleneck which augments with current and future instrument development. EasyDD has achieved its objectives and made significant contributions to scientific research. It can also be used as a model for more mature attempts in the future. EasyDD is currently in use by a number of researchers in a number of academic and research institutions to process high-energy diffraction data. These include data collected by different techniques such as Energy Dispersive Diffraction, Angle Dispersive Diffraction and Computer Aided Tomography. EasyDD has already been used in a number of published studies, and is currently in use by the High Energy X-Ray Imaging Technology project. The software was also used by the author to process and analyse datasets collected from synchrotron radiation facilities. In this regard, the thesis presents novel scientific research involving the use of EasyDD to handle large diffraction datasets in the study of alumina-supported metal oxide catalyst bodies. These data were collected using Tomographic Energy Dispersive Diffraction Imaging and Computer Aided Tomography techniques.
1207.4417	Jingwei Liu	Jingwei Liu, Meizhi Xu	Penalty Constraints and Kernelization of M-Estimation Based Fuzzy C-Means	null	null	null	null	cs.CV stat.CO	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	A framework of M-estimation based fuzzy C-means clustering (MFCM) algorithm is proposed with iterative reweighted least squares (IRLS) algorithm, and penalty constraint and kernelization extensions of MFCM algorithms are also developed. Introducing penalty information to the object functions of MFCM algorithms, the spatially constrained fuzzy C-means (SFCM) is extended to penalty constraints MFCM algorithms(abbr. pMFCM).Substituting the Euclidean distance with kernel method, the MFCM and pMFCM algorithms are extended to kernelized MFCM (abbr. KMFCM) and kernelized pMFCM (abbr.pKMFCM) algorithms. The performances of MFCM, pMFCM, KMFCM and pKMFCM algorithms are evaluated in three tasks: pattern recognition on 10 standard data sets from UCI Machine Learning databases, noise image segmentation performances on a synthetic image, a magnetic resonance brain image (MRI), and image segmentation of a standard images from Berkeley Segmentation Dataset and Benchmark. The experimental results demonstrate the effectiveness of our proposed algorithms in pattern recognition and image segmentation.	[ { "version": "v1", "created": "Wed, 18 Jul 2012 17:20:32 GMT" }, { "version": "v2", "created": "Sat, 19 Jan 2013 10:33:02 GMT" } ]	2013-01-22T00:00:00	[ [ "Liu", "Jingwei", "" ], [ "Xu", "Meizhi", "" ] ]	TITLE: Penalty Constraints and Kernelization of M-Estimation Based Fuzzy C-Means ABSTRACT: A framework of M-estimation based fuzzy C-means clustering (MFCM) algorithm is proposed with iterative reweighted least squares (IRLS) algorithm, and penalty constraint and kernelization extensions of MFCM algorithms are also developed. Introducing penalty information to the object functions of MFCM algorithms, the spatially constrained fuzzy C-means (SFCM) is extended to penalty constraints MFCM algorithms(abbr. pMFCM).Substituting the Euclidean distance with kernel method, the MFCM and pMFCM algorithms are extended to kernelized MFCM (abbr. KMFCM) and kernelized pMFCM (abbr.pKMFCM) algorithms. The performances of MFCM, pMFCM, KMFCM and pKMFCM algorithms are evaluated in three tasks: pattern recognition on 10 standard data sets from UCI Machine Learning databases, noise image segmentation performances on a synthetic image, a magnetic resonance brain image (MRI), and image segmentation of a standard images from Berkeley Segmentation Dataset and Benchmark. The experimental results demonstrate the effectiveness of our proposed algorithms in pattern recognition and image segmentation.
1301.3753	Leif Johnson	Leif Johnson and Craig Corcoran	Switched linear encoding with rectified linear autoencoders	null	null	null	null	cs.LG	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Several recent results in machine learning have established formal connections between autoencoders---artificial neural network models that attempt to reproduce their inputs---and other coding models like sparse coding and K-means. This paper explores in depth an autoencoder model that is constructed using rectified linear activations on its hidden units. Our analysis builds on recent results to further unify the world of sparse linear coding models. We provide an intuitive interpretation of the behavior of these coding models and demonstrate this intuition using small, artificial datasets with known distributions.	[ { "version": "v1", "created": "Wed, 16 Jan 2013 17:04:10 GMT" }, { "version": "v2", "created": "Sat, 19 Jan 2013 19:38:36 GMT" } ]	2013-01-22T00:00:00	[ [ "Johnson", "Leif", "" ], [ "Corcoran", "Craig", "" ] ]	TITLE: Switched linear encoding with rectified linear autoencoders ABSTRACT: Several recent results in machine learning have established formal connections between autoencoders---artificial neural network models that attempt to reproduce their inputs---and other coding models like sparse coding and K-means. This paper explores in depth an autoencoder model that is constructed using rectified linear activations on its hidden units. Our analysis builds on recent results to further unify the world of sparse linear coding models. We provide an intuitive interpretation of the behavior of these coding models and demonstrate this intuition using small, artificial datasets with known distributions.
1301.3844	Gregory F. Cooper	Gregory F. Cooper	A Bayesian Method for Causal Modeling and Discovery Under Selection	Appears in Proceedings of the Sixteenth Conference on Uncertainty in Artificial Intelligence (UAI2000)	null	null	UAI-P-2000-PG-98-106	cs.AI	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	This paper describes a Bayesian method for learning causal networks using samples that were selected in a non-random manner from a population of interest. Examples of data obtained by non-random sampling include convenience samples and case-control data in which a fixed number of samples with and without some condition is collected; such data are not uncommon. The paper describes a method for combining data under selection with prior beliefs in order to derive a posterior probability for a model of the causal processes that are generating the data in the population of interest. The priors include beliefs about the nature of the non-random sampling procedure. Although exact application of the method would be computationally intractable for most realistic datasets, efficient special-case and approximation methods are discussed. Finally, the paper describes how to combine learning under selection with previous methods for learning from observational and experimental data that are obtained on random samples of the population of interest. The net result is a Bayesian methodology that supports causal modeling and discovery from a rich mixture of different types of data.	[ { "version": "v1", "created": "Wed, 16 Jan 2013 15:49:26 GMT" } ]	2013-01-18T00:00:00	[ [ "Cooper", "Gregory F.", "" ] ]	TITLE: A Bayesian Method for Causal Modeling and Discovery Under Selection ABSTRACT: This paper describes a Bayesian method for learning causal networks using samples that were selected in a non-random manner from a population of interest. Examples of data obtained by non-random sampling include convenience samples and case-control data in which a fixed number of samples with and without some condition is collected; such data are not uncommon. The paper describes a method for combining data under selection with prior beliefs in order to derive a posterior probability for a model of the causal processes that are generating the data in the population of interest. The priors include beliefs about the nature of the non-random sampling procedure. Although exact application of the method would be computationally intractable for most realistic datasets, efficient special-case and approximation methods are discussed. Finally, the paper describes how to combine learning under selection with previous methods for learning from observational and experimental data that are obtained on random samples of the population of interest. The net result is a Bayesian methodology that supports causal modeling and discovery from a rich mixture of different types of data.
1301.3856	Nir Friedman	Nir Friedman, Daphne Koller	Being Bayesian about Network Structure	Appears in Proceedings of the Sixteenth Conference on Uncertainty in Artificial Intelligence (UAI2000)	null	null	UAI-P-2000-PG-201-210	cs.LG cs.AI stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	In many domains, we are interested in analyzing the structure of the underlying distribution, e.g., whether one variable is a direct parent of the other. Bayesian model-selection attempts to find the MAP model and use its structure to answer these questions. However, when the amount of available data is modest, there might be many models that have non-negligible posterior. Thus, we want compute the Bayesian posterior of a feature, i.e., the total posterior probability of all models that contain it. In this paper, we propose a new approach for this task. We first show how to efficiently compute a sum over the exponential number of networks that are consistent with a fixed ordering over network variables. This allows us to compute, for a given ordering, both the marginal probability of the data and the posterior of a feature. We then use this result as the basis for an algorithm that approximates the Bayesian posterior of a feature. Our approach uses a Markov Chain Monte Carlo (MCMC) method, but over orderings rather than over network structures. The space of orderings is much smaller and more regular than the space of structures, and has a smoother posterior `landscape'. We present empirical results on synthetic and real-life datasets that compare our approach to full model averaging (when possible), to MCMC over network structures, and to a non-Bayesian bootstrap approach.	[ { "version": "v1", "created": "Wed, 16 Jan 2013 15:50:14 GMT" } ]	2013-01-18T00:00:00	[ [ "Friedman", "Nir", "" ], [ "Koller", "Daphne", "" ] ]	TITLE: Being Bayesian about Network Structure ABSTRACT: In many domains, we are interested in analyzing the structure of the underlying distribution, e.g., whether one variable is a direct parent of the other. Bayesian model-selection attempts to find the MAP model and use its structure to answer these questions. However, when the amount of available data is modest, there might be many models that have non-negligible posterior. Thus, we want compute the Bayesian posterior of a feature, i.e., the total posterior probability of all models that contain it. In this paper, we propose a new approach for this task. We first show how to efficiently compute a sum over the exponential number of networks that are consistent with a fixed ordering over network variables. This allows us to compute, for a given ordering, both the marginal probability of the data and the posterior of a feature. We then use this result as the basis for an algorithm that approximates the Bayesian posterior of a feature. Our approach uses a Markov Chain Monte Carlo (MCMC) method, but over orderings rather than over network structures. The space of orderings is much smaller and more regular than the space of structures, and has a smoother posterior `landscape'. We present empirical results on synthetic and real-life datasets that compare our approach to full model averaging (when possible), to MCMC over network structures, and to a non-Bayesian bootstrap approach.
1301.3884	Dmitry Y. Pavlov	Dmitry Y. Pavlov, Heikki Mannila, Padhraic Smyth	Probabilistic Models for Query Approximation with Large Sparse Binary Datasets	Appears in Proceedings of the Sixteenth Conference on Uncertainty in Artificial Intelligence (UAI2000)	null	null	UAI-P-2000-PG-465-472	cs.AI cs.DB	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Large sparse sets of binary transaction data with millions of records and thousands of attributes occur in various domains: customers purchasing products, users visiting web pages, and documents containing words are just three typical examples. Real-time query selectivity estimation (the problem of estimating the number of rows in the data satisfying a given predicate) is an important practical problem for such databases. We investigate the application of probabilistic models to this problem. In particular, we study a Markov random field (MRF) approach based on frequent sets and maximum entropy, and compare it to the independence model and the Chow-Liu tree model. We find that the MRF model provides substantially more accurate probability estimates than the other methods but is more expensive from a computational and memory viewpoint. To alleviate the computational requirements we show how one can apply bucket elimination and clique tree approaches to take advantage of structure in the models and in the queries. We provide experimental results on two large real-world transaction datasets.	[ { "version": "v1", "created": "Wed, 16 Jan 2013 15:52:06 GMT" } ]	2013-01-18T00:00:00	[ [ "Pavlov", "Dmitry Y.", "" ], [ "Mannila", "Heikki", "" ], [ "Smyth", "Padhraic", "" ] ]	TITLE: Probabilistic Models for Query Approximation with Large Sparse Binary Datasets ABSTRACT: Large sparse sets of binary transaction data with millions of records and thousands of attributes occur in various domains: customers purchasing products, users visiting web pages, and documents containing words are just three typical examples. Real-time query selectivity estimation (the problem of estimating the number of rows in the data satisfying a given predicate) is an important practical problem for such databases. We investigate the application of probabilistic models to this problem. In particular, we study a Markov random field (MRF) approach based on frequent sets and maximum entropy, and compare it to the independence model and the Chow-Liu tree model. We find that the MRF model provides substantially more accurate probability estimates than the other methods but is more expensive from a computational and memory viewpoint. To alleviate the computational requirements we show how one can apply bucket elimination and clique tree approaches to take advantage of structure in the models and in the queries. We provide experimental results on two large real-world transaction datasets.
1301.3891	Marc Sebban	Marc Sebban, Richard Nock	Combining Feature and Prototype Pruning by Uncertainty Minimization	Appears in Proceedings of the Sixteenth Conference on Uncertainty in Artificial Intelligence (UAI2000)	null	null	UAI-P-2000-PG-533-540	cs.LG stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	We focus in this paper on dataset reduction techniques for use in k-nearest neighbor classification. In such a context, feature and prototype selections have always been independently treated by the standard storage reduction algorithms. While this certifying is theoretically justified by the fact that each subproblem is NP-hard, we assume in this paper that a joint storage reduction is in fact more intuitive and can in practice provide better results than two independent processes. Moreover, it avoids a lot of distance calculations by progressively removing useless instances during the feature pruning. While standard selection algorithms often optimize the accuracy to discriminate the set of solutions, we use in this paper a criterion based on an uncertainty measure within a nearest-neighbor graph. This choice comes from recent results that have proven that accuracy is not always the suitable criterion to optimize. In our approach, a feature or an instance is removed if its deletion improves information of the graph. Numerous experiments are presented in this paper and a statistical analysis shows the relevance of our approach, and its tolerance in the presence of noise.	[ { "version": "v1", "created": "Wed, 16 Jan 2013 15:52:33 GMT" } ]	2013-01-18T00:00:00	[ [ "Sebban", "Marc", "" ], [ "Nock", "Richard", "" ] ]	TITLE: Combining Feature and Prototype Pruning by Uncertainty Minimization ABSTRACT: We focus in this paper on dataset reduction techniques for use in k-nearest neighbor classification. In such a context, feature and prototype selections have always been independently treated by the standard storage reduction algorithms. While this certifying is theoretically justified by the fact that each subproblem is NP-hard, we assume in this paper that a joint storage reduction is in fact more intuitive and can in practice provide better results than two independent processes. Moreover, it avoids a lot of distance calculations by progressively removing useless instances during the feature pruning. While standard selection algorithms often optimize the accuracy to discriminate the set of solutions, we use in this paper a criterion based on an uncertainty measure within a nearest-neighbor graph. This choice comes from recent results that have proven that accuracy is not always the suitable criterion to optimize. In our approach, a feature or an instance is removed if its deletion improves information of the graph. Numerous experiments are presented in this paper and a statistical analysis shows the relevance of our approach, and its tolerance in the presence of noise.
1301.4028	Michael Schreiber	Michael Schreiber	Do we need the g-index?	7 pages, 3 figures accepted for publication in Journal of the American Society for Information Science and Technology	null	null	null	physics.soc-ph cs.DL	http://creativecommons.org/licenses/by-nc-sa/3.0/	Using a very small sample of 8 datasets it was recently shown by De Visscher (2011) that the g-index is very close to the square root of the total number of citations. It was argued that there is no bibliometrically meaningful difference. Using another somewhat larger empirical sample of 26 datasets I show that the difference may be larger and I argue in favor of the g-index.	[ { "version": "v1", "created": "Thu, 17 Jan 2013 09:45:27 GMT" } ]	2013-01-18T00:00:00	[ [ "Schreiber", "Michael", "" ] ]	TITLE: Do we need the g-index? ABSTRACT: Using a very small sample of 8 datasets it was recently shown by De Visscher (2011) that the g-index is very close to the square root of the total number of citations. It was argued that there is no bibliometrically meaningful difference. Using another somewhat larger empirical sample of 26 datasets I show that the difference may be larger and I argue in favor of the g-index.
1301.4171	Jason Weston	Jason Weston, Ron Weiss, Hector Yee	Affinity Weighted Embedding	null	null	null	null	cs.IR cs.LG stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Supervised (linear) embedding models like Wsabie and PSI have proven successful at ranking, recommendation and annotation tasks. However, despite being scalable to large datasets they do not take full advantage of the extra data due to their linear nature, and typically underfit. We propose a new class of models which aim to provide improved performance while retaining many of the benefits of the existing class of embedding models. Our new approach works by iteratively learning a linear embedding model where the next iteration's features and labels are reweighted as a function of the previous iteration. We describe several variants of the family, and give some initial results.	[ { "version": "v1", "created": "Thu, 17 Jan 2013 17:46:27 GMT" } ]	2013-01-18T00:00:00	[ [ "Weston", "Jason", "" ], [ "Weiss", "Ron", "" ], [ "Yee", "Hector", "" ] ]	TITLE: Affinity Weighted Embedding ABSTRACT: Supervised (linear) embedding models like Wsabie and PSI have proven successful at ranking, recommendation and annotation tasks. However, despite being scalable to large datasets they do not take full advantage of the extra data due to their linear nature, and typically underfit. We propose a new class of models which aim to provide improved performance while retaining many of the benefits of the existing class of embedding models. Our new approach works by iteratively learning a linear embedding model where the next iteration's features and labels are reweighted as a function of the previous iteration. We describe several variants of the family, and give some initial results.
1207.0166	Claudio Gentile	Claudio Gentile and Francesco Orabona	On Multilabel Classification and Ranking with Partial Feedback	null	null	null	null	cs.LG	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	We present a novel multilabel/ranking algorithm working in partial information settings. The algorithm is based on 2nd-order descent methods, and relies on upper-confidence bounds to trade-off exploration and exploitation. We analyze this algorithm in a partial adversarial setting, where covariates can be adversarial, but multilabel probabilities are ruled by (generalized) linear models. We show O(T^{1/2} log T) regret bounds, which improve in several ways on the existing results. We test the effectiveness of our upper-confidence scheme by contrasting against full-information baselines on real-world multilabel datasets, often obtaining comparable performance.	[ { "version": "v1", "created": "Sat, 30 Jun 2012 23:07:03 GMT" }, { "version": "v2", "created": "Tue, 20 Nov 2012 16:48:22 GMT" }, { "version": "v3", "created": "Wed, 16 Jan 2013 19:19:34 GMT" } ]	2013-01-17T00:00:00	[ [ "Gentile", "Claudio", "" ], [ "Orabona", "Francesco", "" ] ]	TITLE: On Multilabel Classification and Ranking with Partial Feedback ABSTRACT: We present a novel multilabel/ranking algorithm working in partial information settings. The algorithm is based on 2nd-order descent methods, and relies on upper-confidence bounds to trade-off exploration and exploitation. We analyze this algorithm in a partial adversarial setting, where covariates can be adversarial, but multilabel probabilities are ruled by (generalized) linear models. We show O(T^{1/2} log T) regret bounds, which improve in several ways on the existing results. We test the effectiveness of our upper-confidence scheme by contrasting against full-information baselines on real-world multilabel datasets, often obtaining comparable performance.
1301.3528	Momiao Xiong	Momiao Xiong and Long Ma	An Efficient Sufficient Dimension Reduction Method for Identifying Genetic Variants of Clinical Significance	null	null	null	null	q-bio.GN cs.LG stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Fast and cheaper next generation sequencing technologies will generate unprecedentedly massive and highly-dimensional genomic and epigenomic variation data. In the near future, a routine part of medical record will include the sequenced genomes. A fundamental question is how to efficiently extract genomic and epigenomic variants of clinical utility which will provide information for optimal wellness and interference strategies. Traditional paradigm for identifying variants of clinical validity is to test association of the variants. However, significantly associated genetic variants may or may not be usefulness for diagnosis and prognosis of diseases. Alternative to association studies for finding genetic variants of predictive utility is to systematically search variants that contain sufficient information for phenotype prediction. To achieve this, we introduce concepts of sufficient dimension reduction and coordinate hypothesis which project the original high dimensional data to very low dimensional space while preserving all information on response phenotypes. We then formulate clinically significant genetic variant discovery problem into sparse SDR problem and develop algorithms that can select significant genetic variants from up to or even ten millions of predictors with the aid of dividing SDR for whole genome into a number of subSDR problems defined for genomic regions. The sparse SDR is in turn formulated as sparse optimal scoring problem, but with penalty which can remove row vectors from the basis matrix. To speed up computation, we develop the modified alternating direction method for multipliers to solve the sparse optimal scoring problem which can easily be implemented in parallel. To illustrate its application, the proposed method is applied to simulation data and the NHLBI's Exome Sequencing Project dataset	[ { "version": "v1", "created": "Tue, 15 Jan 2013 23:19:14 GMT" } ]	2013-01-17T00:00:00	[ [ "Xiong", "Momiao", "" ], [ "Ma", "Long", "" ] ]	TITLE: An Efficient Sufficient Dimension Reduction Method for Identifying Genetic Variants of Clinical Significance ABSTRACT: Fast and cheaper next generation sequencing technologies will generate unprecedentedly massive and highly-dimensional genomic and epigenomic variation data. In the near future, a routine part of medical record will include the sequenced genomes. A fundamental question is how to efficiently extract genomic and epigenomic variants of clinical utility which will provide information for optimal wellness and interference strategies. Traditional paradigm for identifying variants of clinical validity is to test association of the variants. However, significantly associated genetic variants may or may not be usefulness for diagnosis and prognosis of diseases. Alternative to association studies for finding genetic variants of predictive utility is to systematically search variants that contain sufficient information for phenotype prediction. To achieve this, we introduce concepts of sufficient dimension reduction and coordinate hypothesis which project the original high dimensional data to very low dimensional space while preserving all information on response phenotypes. We then formulate clinically significant genetic variant discovery problem into sparse SDR problem and develop algorithms that can select significant genetic variants from up to or even ten millions of predictors with the aid of dividing SDR for whole genome into a number of subSDR problems defined for genomic regions. The sparse SDR is in turn formulated as sparse optimal scoring problem, but with penalty which can remove row vectors from the basis matrix. To speed up computation, we develop the modified alternating direction method for multipliers to solve the sparse optimal scoring problem which can easily be implemented in parallel. To illustrate its application, the proposed method is applied to simulation data and the NHLBI's Exome Sequencing Project dataset
1301.3539	Yoonseop Kang	Yoonseop Kang and Seungjin Choi	Learning Features with Structure-Adapting Multi-view Exponential Family Harmoniums	3 pages, 2 figures, ICLR2013 workshop track submission	null	null	null	cs.LG	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	We proposea graphical model for multi-view feature extraction that automatically adapts its structure to achieve better representation of data distribution. The proposed model, structure-adapting multi-view harmonium (SA-MVH) has switch parameters that control the connection between hidden nodes and input views, and learn the switch parameter while training. Numerical experiments on synthetic and a real-world dataset demonstrate the useful behavior of the SA-MVH, compared to existing multi-view feature extraction methods.	[ { "version": "v1", "created": "Wed, 16 Jan 2013 01:07:38 GMT" } ]	2013-01-17T00:00:00	[ [ "Kang", "Yoonseop", "" ], [ "Choi", "Seungjin", "" ] ]	TITLE: Learning Features with Structure-Adapting Multi-view Exponential Family Harmoniums ABSTRACT: We proposea graphical model for multi-view feature extraction that automatically adapts its structure to achieve better representation of data distribution. The proposed model, structure-adapting multi-view harmonium (SA-MVH) has switch parameters that control the connection between hidden nodes and input views, and learn the switch parameter while training. Numerical experiments on synthetic and a real-world dataset demonstrate the useful behavior of the SA-MVH, compared to existing multi-view feature extraction methods.
1301.3557	Matthew Zeiler	Matthew D. Zeiler and Rob Fergus	Stochastic Pooling for Regularization of Deep Convolutional Neural Networks	9 pages	null	null	null	cs.LG cs.NE stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	We introduce a simple and effective method for regularizing large convolutional neural networks. We replace the conventional deterministic pooling operations with a stochastic procedure, randomly picking the activation within each pooling region according to a multinomial distribution, given by the activities within the pooling region. The approach is hyper-parameter free and can be combined with other regularization approaches, such as dropout and data augmentation. We achieve state-of-the-art performance on four image datasets, relative to other approaches that do not utilize data augmentation.	[ { "version": "v1", "created": "Wed, 16 Jan 2013 02:12:07 GMT" } ]	2013-01-17T00:00:00	[ [ "Zeiler", "Matthew D.", "" ], [ "Fergus", "Rob", "" ] ]	TITLE: Stochastic Pooling for Regularization of Deep Convolutional Neural Networks ABSTRACT: We introduce a simple and effective method for regularizing large convolutional neural networks. We replace the conventional deterministic pooling operations with a stochastic procedure, randomly picking the activation within each pooling region according to a multinomial distribution, given by the activities within the pooling region. The approach is hyper-parameter free and can be combined with other regularization approaches, such as dropout and data augmentation. We achieve state-of-the-art performance on four image datasets, relative to other approaches that do not utilize data augmentation.
1301.3744	Tim Vines	Timothy H. Vines, Rose L. Andrew, Dan G. Bock, Michelle T. Franklin, Kimberly J. Gilbert, Nolan C. Kane, Jean-S\'ebastien Moore, Brook T. Moyers, S\'ebastien Renaut, Diana J. Rennison, Thor Veen, Sam Yeaman	Mandated data archiving greatly improves access to research data	null	null	10.1096/fj.12-218164	null	cs.DL physics.soc-ph q-bio.QM	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	The data underlying scientific papers should be accessible to researchers both now and in the future, but how best can we ensure that these data are available? Here we examine the effectiveness of four approaches to data archiving: no stated archiving policy, recommending (but not requiring) archiving, and two versions of mandating data deposition at acceptance. We control for differences between data types by trying to obtain data from papers that use a single, widespread population genetic analysis, STRUCTURE. At one extreme, we found that mandated data archiving policies that require the inclusion of a data availability statement in the manuscript improve the odds of finding the data online almost a thousand-fold compared to having no policy. However, archiving rates at journals with less stringent policies were only very slightly higher than those with no policy at all. At one extreme, we found that mandated data archiving policies that require the inclusion of a data availability statement in the manuscript improve the odds of finding the data online almost a thousand fold compared to having no policy. However, archiving rates at journals with less stringent policies were only very slightly higher than those with no policy at all. We also assessed the effectiveness of asking for data directly from authors and obtained over half of the requested datasets, albeit with about 8 days delay and some disagreement with authors. Given the long term benefits of data accessibility to the academic community, we believe that journal based mandatory data archiving policies and mandatory data availability statements should be more widely adopted.	[ { "version": "v1", "created": "Wed, 16 Jan 2013 16:22:26 GMT" } ]	2013-01-17T00:00:00	[ [ "Vines", "Timothy H.", "" ], [ "Andrew", "Rose L.", "" ], [ "Bock", "Dan G.", "" ], [ "Franklin", "Michelle T.", "" ], [ "Gilbert", "Kimberly J.", "" ], [ "Kane", "Nolan C.", "" ], [ "Moore", "Jean-Sébastien", "" ], [ "Moyers", "Brook T.", "" ], [ "Renaut", "Sébastien", "" ], [ "Rennison", "Diana J.", "" ], [ "Veen", "Thor", "" ], [ "Yeaman", "Sam", "" ] ]	TITLE: Mandated data archiving greatly improves access to research data ABSTRACT: The data underlying scientific papers should be accessible to researchers both now and in the future, but how best can we ensure that these data are available? Here we examine the effectiveness of four approaches to data archiving: no stated archiving policy, recommending (but not requiring) archiving, and two versions of mandating data deposition at acceptance. We control for differences between data types by trying to obtain data from papers that use a single, widespread population genetic analysis, STRUCTURE. At one extreme, we found that mandated data archiving policies that require the inclusion of a data availability statement in the manuscript improve the odds of finding the data online almost a thousand-fold compared to having no policy. However, archiving rates at journals with less stringent policies were only very slightly higher than those with no policy at all. At one extreme, we found that mandated data archiving policies that require the inclusion of a data availability statement in the manuscript improve the odds of finding the data online almost a thousand fold compared to having no policy. However, archiving rates at journals with less stringent policies were only very slightly higher than those with no policy at all. We also assessed the effectiveness of asking for data directly from authors and obtained over half of the requested datasets, albeit with about 8 days delay and some disagreement with authors. Given the long term benefits of data accessibility to the academic community, we believe that journal based mandatory data archiving policies and mandatory data availability statements should be more widely adopted.
1301.2659	Fabrice Rossi	Romain Guigour\`es, Marc Boull\'e, Fabrice Rossi (SAMM)	A Triclustering Approach for Time Evolving Graphs	null	Co-clustering and Applications International Conference on Data Mining Workshop, Brussels : Belgium (2012)	10.1109/ICDMW.2012.61	null	cs.LG cs.SI stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	This paper introduces a novel technique to track structures in time evolving graphs. The method is based on a parameter free approach for three-dimensional co-clustering of the source vertices, the target vertices and the time. All these features are simultaneously segmented in order to build time segments and clusters of vertices whose edge distributions are similar and evolve in the same way over the time segments. The main novelty of this approach lies in that the time segments are directly inferred from the evolution of the edge distribution between the vertices, thus not requiring the user to make an a priori discretization. Experiments conducted on a synthetic dataset illustrate the good behaviour of the technique, and a study of a real-life dataset shows the potential of the proposed approach for exploratory data analysis.	[ { "version": "v1", "created": "Sat, 12 Jan 2013 07:51:14 GMT" } ]	2013-01-15T00:00:00	[ [ "Guigourès", "Romain", "", "SAMM" ], [ "Boullé", "Marc", "", "SAMM" ], [ "Rossi", "Fabrice", "", "SAMM" ] ]	TITLE: A Triclustering Approach for Time Evolving Graphs ABSTRACT: This paper introduces a novel technique to track structures in time evolving graphs. The method is based on a parameter free approach for three-dimensional co-clustering of the source vertices, the target vertices and the time. All these features are simultaneously segmented in order to build time segments and clusters of vertices whose edge distributions are similar and evolve in the same way over the time segments. The main novelty of this approach lies in that the time segments are directly inferred from the evolution of the edge distribution between the vertices, thus not requiring the user to make an a priori discretization. Experiments conducted on a synthetic dataset illustrate the good behaviour of the technique, and a study of a real-life dataset shows the potential of the proposed approach for exploratory data analysis.
1301.2785	Rafi Muhammad	Muhammad Rafi, Mohammad Shahid Shaikh	A comparison of SVM and RVM for Document Classification	ICoCSIM 2012, Medan Indonesia	null	null	null	cs.IR cs.LG	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Document classification is a task of assigning a new unclassified document to one of the predefined set of classes. The content based document classification uses the content of the document with some weighting criteria to assign it to one of the predefined classes. It is a major task in library science, electronic document management systems and information sciences. This paper investigates document classification by using two different classification techniques (1) Support Vector Machine (SVM) and (2) Relevance Vector Machine (RVM). SVM is a supervised machine learning technique that can be used for classification task. In its basic form, SVM represents the instances of the data into space and tries to separate the distinct classes by a maximum possible wide gap (hyper plane) that separates the classes. On the other hand RVM uses probabilistic measure to define this separation space. RVM uses Bayesian inference to obtain succinct solution, thus RVM uses significantly fewer basis functions. Experimental studies on three standard text classification datasets reveal that although RVM takes more training time, its classification is much better as compared to SVM.	[ { "version": "v1", "created": "Sun, 13 Jan 2013 15:58:09 GMT" } ]	2013-01-15T00:00:00	[ [ "Rafi", "Muhammad", "" ], [ "Shaikh", "Mohammad Shahid", "" ] ]	TITLE: A comparison of SVM and RVM for Document Classification ABSTRACT: Document classification is a task of assigning a new unclassified document to one of the predefined set of classes. The content based document classification uses the content of the document with some weighting criteria to assign it to one of the predefined classes. It is a major task in library science, electronic document management systems and information sciences. This paper investigates document classification by using two different classification techniques (1) Support Vector Machine (SVM) and (2) Relevance Vector Machine (RVM). SVM is a supervised machine learning technique that can be used for classification task. In its basic form, SVM represents the instances of the data into space and tries to separate the distinct classes by a maximum possible wide gap (hyper plane) that separates the classes. On the other hand RVM uses probabilistic measure to define this separation space. RVM uses Bayesian inference to obtain succinct solution, thus RVM uses significantly fewer basis functions. Experimental studies on three standard text classification datasets reveal that although RVM takes more training time, its classification is much better as compared to SVM.
1301.2283	Tomas Kocka	Tomas Kocka, Robert Castelo	Improved learning of Bayesian networks	Appears in Proceedings of the Seventeenth Conference on Uncertainty in Artificial Intelligence (UAI2001)	null	null	UAI-P-2001-PG-269-276	cs.LG cs.AI stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	The search space of Bayesian Network structures is usually defined as Acyclic Directed Graphs (DAGs) and the search is done by local transformations of DAGs. But the space of Bayesian Networks is ordered by DAG Markov model inclusion and it is natural to consider that a good search policy should take this into account. First attempt to do this (Chickering 1996) was using equivalence classes of DAGs instead of DAGs itself. This approach produces better results but it is significantly slower. We present a compromise between these two approaches. It uses DAGs to search the space in such a way that the ordering by inclusion is taken into account. This is achieved by repetitive usage of local moves within the equivalence class of DAGs. We show that this new approach produces better results than the original DAGs approach without substantial change in time complexity. We present empirical results, within the framework of heuristic search and Markov Chain Monte Carlo, provided through the Alarm dataset.	[ { "version": "v1", "created": "Thu, 10 Jan 2013 16:24:32 GMT" } ]	2013-01-14T00:00:00	[ [ "Kocka", "Tomas", "" ], [ "Castelo", "Robert", "" ] ]	TITLE: Improved learning of Bayesian networks ABSTRACT: The search space of Bayesian Network structures is usually defined as Acyclic Directed Graphs (DAGs) and the search is done by local transformations of DAGs. But the space of Bayesian Networks is ordered by DAG Markov model inclusion and it is natural to consider that a good search policy should take this into account. First attempt to do this (Chickering 1996) was using equivalence classes of DAGs instead of DAGs itself. This approach produces better results but it is significantly slower. We present a compromise between these two approaches. It uses DAGs to search the space in such a way that the ordering by inclusion is taken into account. This is achieved by repetitive usage of local moves within the equivalence class of DAGs. We show that this new approach produces better results than the original DAGs approach without substantial change in time complexity. We present empirical results, within the framework of heuristic search and Markov Chain Monte Carlo, provided through the Alarm dataset.
1301.2375	Jianxin Li	Jianxin Li, Chengfei Liu, Liang Yao and Jeffrey Xu Yu	Context-based Diversification for Keyword Queries over XML Data	null	null	null	null	cs.DB	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	While keyword query empowers ordinary users to search vast amount of data, the ambiguity of keyword query makes it difficult to effectively answer keyword queries, especially for short and vague keyword queries. To address this challenging problem, in this paper we propose an approach that automatically diversifies XML keyword search based on its different contexts in the XML data. Given a short and vague keyword query and XML data to be searched, we firstly derive keyword search candidates of the query by a classifical feature selection model. And then, we design an effective XML keyword search diversification model to measure the quality of each candidate. After that, three efficient algorithms are proposed to evaluate the possible generated query candidates representing the diversified search intentions, from which we can find and return top-$k$ qualified query candidates that are most relevant to the given keyword query while they can cover maximal number of distinct results.At last, a comprehensive evaluation on real and synthetic datasets demonstrates the effectiveness of our proposed diversification model and the efficiency of our algorithms.	[ { "version": "v1", "created": "Fri, 11 Jan 2013 01:33:50 GMT" } ]	2013-01-14T00:00:00	[ [ "Li", "Jianxin", "" ], [ "Liu", "Chengfei", "" ], [ "Yao", "Liang", "" ], [ "Yu", "Jeffrey Xu", "" ] ]	TITLE: Context-based Diversification for Keyword Queries over XML Data ABSTRACT: While keyword query empowers ordinary users to search vast amount of data, the ambiguity of keyword query makes it difficult to effectively answer keyword queries, especially for short and vague keyword queries. To address this challenging problem, in this paper we propose an approach that automatically diversifies XML keyword search based on its different contexts in the XML data. Given a short and vague keyword query and XML data to be searched, we firstly derive keyword search candidates of the query by a classifical feature selection model. And then, we design an effective XML keyword search diversification model to measure the quality of each candidate. After that, three efficient algorithms are proposed to evaluate the possible generated query candidates representing the diversified search intentions, from which we can find and return top-$k$ qualified query candidates that are most relevant to the given keyword query while they can cover maximal number of distinct results.At last, a comprehensive evaluation on real and synthetic datasets demonstrates the effectiveness of our proposed diversification model and the efficiency of our algorithms.
1301.2378	Jianxin Li	Jianxin Li, Chengfei Liu, Liang Yao, Jeffrey Xu Yu and Rui Zhou	Query-driven Frequent Co-occurring Term Extraction over Relational Data using MapReduce	null	null	null	null	cs.DB	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	In this paper we study how to efficiently compute \textit{frequent co-occurring terms} (FCT) in the results of a keyword query in parallel using the popular MapReduce framework. Taking as input a keyword query q and an integer k, an FCT query reports the k terms that are not in q, but appear most frequently in the results of the keyword query q over multiple joined relations. The returned terms of FCT search can be used to do query expansion and query refinement for traditional keyword search. Different from the method of FCT search in a single platform, our proposed approach can efficiently answer a FCT query using the MapReduce Paradigm without pre-computing the results of the original keyword query, which is run in parallel platform. In this work, we can output the final FCT search results by two MapReduce jobs: the first is to extract the statistical information of the data; and the second is to calculate the total frequency of each term based on the output of the first job. At the two MapReduce jobs, we would guarantee the load balance of mappers and the computational balance of reducers as much as possible. Analytical and experimental evaluations demonstrate the efficiency and scalability of our proposed approach using TPC-H benchmark datasets with different sizes.	[ { "version": "v1", "created": "Fri, 11 Jan 2013 01:55:10 GMT" } ]	2013-01-14T00:00:00	[ [ "Li", "Jianxin", "" ], [ "Liu", "Chengfei", "" ], [ "Yao", "Liang", "" ], [ "Yu", "Jeffrey Xu", "" ], [ "Zhou", "Rui", "" ] ]	TITLE: Query-driven Frequent Co-occurring Term Extraction over Relational Data using MapReduce ABSTRACT: In this paper we study how to efficiently compute \textit{frequent co-occurring terms} (FCT) in the results of a keyword query in parallel using the popular MapReduce framework. Taking as input a keyword query q and an integer k, an FCT query reports the k terms that are not in q, but appear most frequently in the results of the keyword query q over multiple joined relations. The returned terms of FCT search can be used to do query expansion and query refinement for traditional keyword search. Different from the method of FCT search in a single platform, our proposed approach can efficiently answer a FCT query using the MapReduce Paradigm without pre-computing the results of the original keyword query, which is run in parallel platform. In this work, we can output the final FCT search results by two MapReduce jobs: the first is to extract the statistical information of the data; and the second is to calculate the total frequency of each term based on the output of the first job. At the two MapReduce jobs, we would guarantee the load balance of mappers and the computational balance of reducers as much as possible. Analytical and experimental evaluations demonstrate the efficiency and scalability of our proposed approach using TPC-H benchmark datasets with different sizes.
1301.2115	Krikamol Muandet	Krikamol Muandet, David Balduzzi, Bernhard Sch\"olkopf	Domain Generalization via Invariant Feature Representation	The 30th International Conference on Machine Learning (ICML 2013)	null	null	null	stat.ML cs.LG	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	This paper investigates domain generalization: How to take knowledge acquired from an arbitrary number of related domains and apply it to previously unseen domains? We propose Domain-Invariant Component Analysis (DICA), a kernel-based optimization algorithm that learns an invariant transformation by minimizing the dissimilarity across domains, whilst preserving the functional relationship between input and output variables. A learning-theoretic analysis shows that reducing dissimilarity improves the expected generalization ability of classifiers on new domains, motivating the proposed algorithm. Experimental results on synthetic and real-world datasets demonstrate that DICA successfully learns invariant features and improves classifier performance in practice.	[ { "version": "v1", "created": "Thu, 10 Jan 2013 13:29:17 GMT" } ]	2013-01-11T00:00:00	[ [ "Muandet", "Krikamol", "" ], [ "Balduzzi", "David", "" ], [ "Schölkopf", "Bernhard", "" ] ]	TITLE: Domain Generalization via Invariant Feature Representation ABSTRACT: This paper investigates domain generalization: How to take knowledge acquired from an arbitrary number of related domains and apply it to previously unseen domains? We propose Domain-Invariant Component Analysis (DICA), a kernel-based optimization algorithm that learns an invariant transformation by minimizing the dissimilarity across domains, whilst preserving the functional relationship between input and output variables. A learning-theoretic analysis shows that reducing dissimilarity improves the expected generalization ability of classifiers on new domains, motivating the proposed algorithm. Experimental results on synthetic and real-world datasets demonstrate that DICA successfully learns invariant features and improves classifier performance in practice.
1301.1722	Andrea Montanari	Yash Deshpande and Andrea Montanari	Linear Bandits in High Dimension and Recommendation Systems	21 pages, 4 figures	null	null	null	cs.LG stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	A large number of online services provide automated recommendations to help users to navigate through a large collection of items. New items (products, videos, songs, advertisements) are suggested on the basis of the user's past history and --when available-- her demographic profile. Recommendations have to satisfy the dual goal of helping the user to explore the space of available items, while allowing the system to probe the user's preferences. We model this trade-off using linearly parametrized multi-armed bandits, propose a policy and prove upper and lower bounds on the cumulative "reward" that coincide up to constants in the data poor (high-dimensional) regime. Prior work on linear bandits has focused on the data rich (low-dimensional) regime and used cumulative "risk" as the figure of merit. For this data rich regime, we provide a simple modification for our policy that achieves near-optimal risk performance under more restrictive assumptions on the geometry of the problem. We test (a variation of) the scheme used for establishing achievability on the Netflix and MovieLens datasets and obtain good agreement with the qualitative predictions of the theory we develop.	[ { "version": "v1", "created": "Tue, 8 Jan 2013 23:45:06 GMT" } ]	2013-01-10T00:00:00	[ [ "Deshpande", "Yash", "" ], [ "Montanari", "Andrea", "" ] ]	TITLE: Linear Bandits in High Dimension and Recommendation Systems ABSTRACT: A large number of online services provide automated recommendations to help users to navigate through a large collection of items. New items (products, videos, songs, advertisements) are suggested on the basis of the user's past history and --when available-- her demographic profile. Recommendations have to satisfy the dual goal of helping the user to explore the space of available items, while allowing the system to probe the user's preferences. We model this trade-off using linearly parametrized multi-armed bandits, propose a policy and prove upper and lower bounds on the cumulative "reward" that coincide up to constants in the data poor (high-dimensional) regime. Prior work on linear bandits has focused on the data rich (low-dimensional) regime and used cumulative "risk" as the figure of merit. For this data rich regime, we provide a simple modification for our policy that achieves near-optimal risk performance under more restrictive assumptions on the geometry of the problem. We test (a variation of) the scheme used for establishing achievability on the Netflix and MovieLens datasets and obtain good agreement with the qualitative predictions of the theory we develop.
1301.1502	Hannah Inbarani	N. Kalaiselvi, H. Hannah Inbarani	Fuzzy Soft Set Based Classification for Gene Expression Data	7 pages, IJSER Vol.3 Issue: 10 Oct 2012	null	null	null	cs.AI cs.CE	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Classification is one of the major issues in Data Mining Research fields. The classification problems in medical area often classify medical dataset based on the result of medical diagnosis or description of medical treatment by the medical practitioner. This research work discusses the classification process of Gene Expression data for three different cancers which are breast cancer, lung cancer and leukemia cancer with two classes which are cancerous stage and non cancerous stage. We have applied a fuzzy soft set similarity based classifier to enhance the accuracy to predict the stages among cancer genes and the informative genes are selected by using Entopy filtering.	[ { "version": "v1", "created": "Tue, 8 Jan 2013 11:48:49 GMT" } ]	2013-01-09T00:00:00	[ [ "Kalaiselvi", "N.", "" ], [ "Inbarani", "H. Hannah", "" ] ]	TITLE: Fuzzy Soft Set Based Classification for Gene Expression Data ABSTRACT: Classification is one of the major issues in Data Mining Research fields. The classification problems in medical area often classify medical dataset based on the result of medical diagnosis or description of medical treatment by the medical practitioner. This research work discusses the classification process of Gene Expression data for three different cancers which are breast cancer, lung cancer and leukemia cancer with two classes which are cancerous stage and non cancerous stage. We have applied a fuzzy soft set similarity based classifier to enhance the accuracy to predict the stages among cancer genes and the informative genes are selected by using Entopy filtering.
1212.4775	Mario Frank	Mario Frank, Joachim M. Buhmann, David Basin	Role Mining with Probabilistic Models	accepted for publication at ACM Transactions on Information and System Security (TISSEC)	null	null	null	cs.CR cs.LG stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Role mining tackles the problem of finding a role-based access control (RBAC) configuration, given an access-control matrix assigning users to access permissions as input. Most role mining approaches work by constructing a large set of candidate roles and use a greedy selection strategy to iteratively pick a small subset such that the differences between the resulting RBAC configuration and the access control matrix are minimized. In this paper, we advocate an alternative approach that recasts role mining as an inference problem rather than a lossy compression problem. Instead of using combinatorial algorithms to minimize the number of roles needed to represent the access-control matrix, we derive probabilistic models to learn the RBAC configuration that most likely underlies the given matrix. Our models are generative in that they reflect the way that permissions are assigned to users in a given RBAC configuration. We additionally model how user-permission assignments that conflict with an RBAC configuration emerge and we investigate the influence of constraints on role hierarchies and on the number of assignments. In experiments with access-control matrices from real-world enterprises, we compare our proposed models with other role mining methods. Our results show that our probabilistic models infer roles that generalize well to new system users for a wide variety of data, while other models' generalization abilities depend on the dataset given.	[ { "version": "v1", "created": "Wed, 19 Dec 2012 18:12:34 GMT" }, { "version": "v2", "created": "Thu, 3 Jan 2013 17:27:55 GMT" }, { "version": "v3", "created": "Fri, 4 Jan 2013 22:24:15 GMT" } ]	2013-01-08T00:00:00	[ [ "Frank", "Mario", "" ], [ "Buhmann", "Joachim M.", "" ], [ "Basin", "David", "" ] ]	TITLE: Role Mining with Probabilistic Models ABSTRACT: Role mining tackles the problem of finding a role-based access control (RBAC) configuration, given an access-control matrix assigning users to access permissions as input. Most role mining approaches work by constructing a large set of candidate roles and use a greedy selection strategy to iteratively pick a small subset such that the differences between the resulting RBAC configuration and the access control matrix are minimized. In this paper, we advocate an alternative approach that recasts role mining as an inference problem rather than a lossy compression problem. Instead of using combinatorial algorithms to minimize the number of roles needed to represent the access-control matrix, we derive probabilistic models to learn the RBAC configuration that most likely underlies the given matrix. Our models are generative in that they reflect the way that permissions are assigned to users in a given RBAC configuration. We additionally model how user-permission assignments that conflict with an RBAC configuration emerge and we investigate the influence of constraints on role hierarchies and on the number of assignments. In experiments with access-control matrices from real-world enterprises, we compare our proposed models with other role mining methods. Our results show that our probabilistic models infer roles that generalize well to new system users for a wide variety of data, while other models' generalization abilities depend on the dataset given.
1301.0561	David Maxwell Chickering	David Maxwell Chickering, Christopher Meek	Finding Optimal Bayesian Networks	Appears in Proceedings of the Eighteenth Conference on Uncertainty in Artificial Intelligence (UAI2002)	null	null	UAI-P-2002-PG-94-102	cs.AI	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	In this paper, we derive optimality results for greedy Bayesian-network search algorithms that perform single-edge modifications at each step and use asymptotically consistent scoring criteria. Our results extend those of Meek (1997) and Chickering (2002), who demonstrate that in the limit of large datasets, if the generative distribution is perfect with respect to a DAG defined over the observable variables, such search algorithms will identify this optimal (i.e. generative) DAG model. We relax their assumption about the generative distribution, and assume only that this distribution satisfies the {em composition property} over the observable variables, which is a more realistic assumption for real domains. Under this assumption, we guarantee that the search algorithms identify an {em inclusion-optimal} model; that is, a model that (1) contains the generative distribution and (2) has no sub-model that contains this distribution. In addition, we show that the composition property is guaranteed to hold whenever the dependence relationships in the generative distribution can be characterized by paths between singleton elements in some generative graphical model (e.g. a DAG, a chain graph, or a Markov network) even when the generative model includes unobserved variables, and even when the observed data is subject to selection bias.	[ { "version": "v1", "created": "Wed, 12 Dec 2012 15:55:46 GMT" } ]	2013-01-07T00:00:00	[ [ "Chickering", "David Maxwell", "" ], [ "Meek", "Christopher", "" ] ]	TITLE: Finding Optimal Bayesian Networks ABSTRACT: In this paper, we derive optimality results for greedy Bayesian-network search algorithms that perform single-edge modifications at each step and use asymptotically consistent scoring criteria. Our results extend those of Meek (1997) and Chickering (2002), who demonstrate that in the limit of large datasets, if the generative distribution is perfect with respect to a DAG defined over the observable variables, such search algorithms will identify this optimal (i.e. generative) DAG model. We relax their assumption about the generative distribution, and assume only that this distribution satisfies the {em composition property} over the observable variables, which is a more realistic assumption for real domains. Under this assumption, we guarantee that the search algorithms identify an {em inclusion-optimal} model; that is, a model that (1) contains the generative distribution and (2) has no sub-model that contains this distribution. In addition, we show that the composition property is guaranteed to hold whenever the dependence relationships in the generative distribution can be characterized by paths between singleton elements in some generative graphical model (e.g. a DAG, a chain graph, or a Markov network) even when the generative model includes unobserved variables, and even when the observed data is subject to selection bias.
1301.0432	Fahad Mahmood Mr	F. Mahmood, F. Kunwar	A Self-Organizing Neural Scheme for Door Detection in Different Environments	Page No 13-18, 7 figures, Published with International Journal of Computer Applications (IJCA)	International Journal of Computer Applications 60(9):13-18, 2012	10.5120/9719-3679	null	cs.CV	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Doors are important landmarks for indoor mobile robot navigation and also assist blind people to independently access unfamiliar buildings. Most existing algorithms of door detection are limited to work for familiar environments because of restricted assumptions about color, texture and shape. In this paper we propose a novel approach which employs feature based classification and uses the Kohonen Self-Organizing Map (SOM) for the purpose of door detection. Generic and stable features are used for the training of SOM that increase the performance significantly: concavity, bottom-edge intensity profile and door edges. To validate the robustness and generalizability of our method, we collected a large dataset of real world door images from a variety of environments and different lighting conditions. The algorithm achieves more than 95% detection which demonstrates that our door detection method is generic and robust with variations of color, texture, occlusions, lighting condition, scales, and viewpoints.	[ { "version": "v1", "created": "Thu, 3 Jan 2013 12:04:28 GMT" } ]	2013-01-04T00:00:00	[ [ "Mahmood", "F.", "" ], [ "Kunwar", "F.", "" ] ]	TITLE: A Self-Organizing Neural Scheme for Door Detection in Different Environments ABSTRACT: Doors are important landmarks for indoor mobile robot navigation and also assist blind people to independently access unfamiliar buildings. Most existing algorithms of door detection are limited to work for familiar environments because of restricted assumptions about color, texture and shape. In this paper we propose a novel approach which employs feature based classification and uses the Kohonen Self-Organizing Map (SOM) for the purpose of door detection. Generic and stable features are used for the training of SOM that increase the performance significantly: concavity, bottom-edge intensity profile and door edges. To validate the robustness and generalizability of our method, we collected a large dataset of real world door images from a variety of environments and different lighting conditions. The algorithm achieves more than 95% detection which demonstrates that our door detection method is generic and robust with variations of color, texture, occlusions, lighting condition, scales, and viewpoints.
0911.2942	Chris Giannella	Chris Giannella, Kun Liu, Hillol Kargupta	Breaching Euclidean Distance-Preserving Data Perturbation Using Few Known Inputs	This is a major revision accounting for journal peer-review. Changes include: removal of known sample attack, more citations added, an empirical comparison against the algorithm of Kaplan et al. added	Data & Knowledge Engineering 83, pages 93-110, 2013	null	null	cs.DB cs.CR	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	We examine Euclidean distance-preserving data perturbation as a tool for privacy-preserving data mining. Such perturbations allow many important data mining algorithms e.g. hierarchical and k-means clustering), with only minor modification, to be applied to the perturbed data and produce exactly the same results as if applied to the original data. However, the issue of how well the privacy of the original data is preserved needs careful study. We engage in this study by assuming the role of an attacker armed with a small set of known original data tuples (inputs). Little work has been done examining this kind of attack when the number of known original tuples is less than the number of data dimensions. We focus on this important case, develop and rigorously analyze an attack that utilizes any number of known original tuples. The approach allows the attacker to estimate the original data tuple associated with each perturbed tuple and calculate the probability that the estimation results in a privacy breach. On a real 16-dimensional dataset, we show that the attacker, with 4 known original tuples, can estimate an original unknown tuple with less than 7% error with probability exceeding 0.8.	[ { "version": "v1", "created": "Mon, 16 Nov 2009 02:51:37 GMT" }, { "version": "v2", "created": "Wed, 2 Jan 2013 15:49:10 GMT" } ]	2013-01-03T00:00:00	[ [ "Giannella", "Chris", "" ], [ "Liu", "Kun", "" ], [ "Kargupta", "Hillol", "" ] ]	TITLE: Breaching Euclidean Distance-Preserving Data Perturbation Using Few Known Inputs ABSTRACT: We examine Euclidean distance-preserving data perturbation as a tool for privacy-preserving data mining. Such perturbations allow many important data mining algorithms e.g. hierarchical and k-means clustering), with only minor modification, to be applied to the perturbed data and produce exactly the same results as if applied to the original data. However, the issue of how well the privacy of the original data is preserved needs careful study. We engage in this study by assuming the role of an attacker armed with a small set of known original data tuples (inputs). Little work has been done examining this kind of attack when the number of known original tuples is less than the number of data dimensions. We focus on this important case, develop and rigorously analyze an attack that utilizes any number of known original tuples. The approach allows the attacker to estimate the original data tuple associated with each perturbed tuple and calculate the probability that the estimation results in a privacy breach. On a real 16-dimensional dataset, we show that the attacker, with 4 known original tuples, can estimate an original unknown tuple with less than 7% error with probability exceeding 0.8.
1212.5841	Andrei Zinovyev Dr.	Andrei Zinovyev and Evgeny Mirkes	Data complexity measured by principal graphs	Computers and Mathematics with Applications, in press	null	10.1016/j.camwa.2012.12.009	null	cs.LG cs.IT math.IT	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	How to measure the complexity of a finite set of vectors embedded in a multidimensional space? This is a non-trivial question which can be approached in many different ways. Here we suggest a set of data complexity measures using universal approximators, principal cubic complexes. Principal cubic complexes generalise the notion of principal manifolds for datasets with non-trivial topologies. The type of the principal cubic complex is determined by its dimension and a grammar of elementary graph transformations. The simplest grammar produces principal trees. We introduce three natural types of data complexity: 1) geometric (deviation of the data's approximator from some "idealized" configuration, such as deviation from harmonicity); 2) structural (how many elements of a principal graph are needed to approximate the data), and 3) construction complexity (how many applications of elementary graph transformations are needed to construct the principal object starting from the simplest one). We compute these measures for several simulated and real-life data distributions and show them in the "accuracy-complexity" plots, helping to optimize the accuracy/complexity ratio. We discuss various issues connected with measuring data complexity. Software for computing data complexity measures from principal cubic complexes is provided as well.	[ { "version": "v1", "created": "Sun, 23 Dec 2012 23:20:14 GMT" }, { "version": "v2", "created": "Wed, 2 Jan 2013 00:00:40 GMT" } ]	2013-01-03T00:00:00	[ [ "Zinovyev", "Andrei", "" ], [ "Mirkes", "Evgeny", "" ] ]	TITLE: Data complexity measured by principal graphs ABSTRACT: How to measure the complexity of a finite set of vectors embedded in a multidimensional space? This is a non-trivial question which can be approached in many different ways. Here we suggest a set of data complexity measures using universal approximators, principal cubic complexes. Principal cubic complexes generalise the notion of principal manifolds for datasets with non-trivial topologies. The type of the principal cubic complex is determined by its dimension and a grammar of elementary graph transformations. The simplest grammar produces principal trees. We introduce three natural types of data complexity: 1) geometric (deviation of the data's approximator from some "idealized" configuration, such as deviation from harmonicity); 2) structural (how many elements of a principal graph are needed to approximate the data), and 3) construction complexity (how many applications of elementary graph transformations are needed to construct the principal object starting from the simplest one). We compute these measures for several simulated and real-life data distributions and show them in the "accuracy-complexity" plots, helping to optimize the accuracy/complexity ratio. We discuss various issues connected with measuring data complexity. Software for computing data complexity measures from principal cubic complexes is provided as well.
1212.6316	Nathalie Villa-Vialaneix	Madalina Olteanu (SAMM), Nathalie Villa-Vialaneix (SAMM), Marie Cottrell (SAMM)	On-line relational SOM for dissimilarity data	WSOM 2012, Santiago : Chile (2012)	null	null	null	stat.ML cs.LG	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	In some applications and in order to address real world situations better, data may be more complex than simple vectors. In some examples, they can be known through their pairwise dissimilarities only. Several variants of the Self Organizing Map algorithm were introduced to generalize the original algorithm to this framework. Whereas median SOM is based on a rough representation of the prototypes, relational SOM allows representing these prototypes by a virtual combination of all elements in the data set. However, this latter approach suffers from two main drawbacks. First, its complexity can be large. Second, only a batch version of this algorithm has been studied so far and it often provides results having a bad topographic organization. In this article, an on-line version of relational SOM is described and justified. The algorithm is tested on several datasets, including categorical data and graphs, and compared with the batch version and with other SOM algorithms for non vector data.	[ { "version": "v1", "created": "Thu, 27 Dec 2012 07:07:06 GMT" } ]	2013-01-03T00:00:00	[ [ "Olteanu", "Madalina", "", "SAMM" ], [ "Villa-Vialaneix", "Nathalie", "", "SAMM" ], [ "Cottrell", "Marie", "", "SAMM" ] ]	TITLE: On-line relational SOM for dissimilarity data ABSTRACT: In some applications and in order to address real world situations better, data may be more complex than simple vectors. In some examples, they can be known through their pairwise dissimilarities only. Several variants of the Self Organizing Map algorithm were introduced to generalize the original algorithm to this framework. Whereas median SOM is based on a rough representation of the prototypes, relational SOM allows representing these prototypes by a virtual combination of all elements in the data set. However, this latter approach suffers from two main drawbacks. First, its complexity can be large. Second, only a batch version of this algorithm has been studied so far and it often provides results having a bad topographic organization. In this article, an on-line version of relational SOM is described and justified. The algorithm is tested on several datasets, including categorical data and graphs, and compared with the batch version and with other SOM algorithms for non vector data.
1301.0082	F. Ozgur Catak	F. Ozgur Catak and M. Erdal Balaban	CloudSVM : Training an SVM Classifier in Cloud Computing Systems	13 pages	null	null	null	cs.LG cs.DC	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	In conventional method, distributed support vector machines (SVM) algorithms are trained over pre-configured intranet/internet environments to find out an optimal classifier. These methods are very complicated and costly for large datasets. Hence, we propose a method that is referred as the Cloud SVM training mechanism (CloudSVM) in a cloud computing environment with MapReduce technique for distributed machine learning applications. Accordingly, (i) SVM algorithm is trained in distributed cloud storage servers that work concurrently; (ii) merge all support vectors in every trained cloud node; and (iii) iterate these two steps until the SVM converges to the optimal classifier function. Large scale data sets are not possible to train using SVM algorithm on a single computer. The results of this study are important for training of large scale data sets for machine learning applications. We provided that iterative training of splitted data set in cloud computing environment using SVM will converge to a global optimal classifier in finite iteration size.	[ { "version": "v1", "created": "Tue, 1 Jan 2013 13:20:27 GMT" } ]	2013-01-03T00:00:00	[ [ "Catak", "F. Ozgur", "" ], [ "Balaban", "M. Erdal", "" ] ]	TITLE: CloudSVM : Training an SVM Classifier in Cloud Computing Systems ABSTRACT: In conventional method, distributed support vector machines (SVM) algorithms are trained over pre-configured intranet/internet environments to find out an optimal classifier. These methods are very complicated and costly for large datasets. Hence, we propose a method that is referred as the Cloud SVM training mechanism (CloudSVM) in a cloud computing environment with MapReduce technique for distributed machine learning applications. Accordingly, (i) SVM algorithm is trained in distributed cloud storage servers that work concurrently; (ii) merge all support vectors in every trained cloud node; and (iii) iterate these two steps until the SVM converges to the optimal classifier function. Large scale data sets are not possible to train using SVM algorithm on a single computer. The results of this study are important for training of large scale data sets for machine learning applications. We provided that iterative training of splitted data set in cloud computing environment using SVM will converge to a global optimal classifier in finite iteration size.
1301.0289	Aaditya Prakash	Aaditya Prakash	Reconstructing Self Organizing Maps as Spider Graphs for better visual interpretation of large unstructured datasets	9 pages, 8 figures	null	null	null	cs.GR stat.ML	http://creativecommons.org/licenses/by/3.0/	Self-Organizing Maps (SOM) are popular unsupervised artificial neural network used to reduce dimensions and visualize data. Visual interpretation from Self-Organizing Maps (SOM) has been limited due to grid approach of data representation, which makes inter-scenario analysis impossible. The paper proposes a new way to structure SOM. This model reconstructs SOM to show strength between variables as the threads of a cobweb and illuminate inter-scenario analysis. While Radar Graphs are very crude representation of spider web, this model uses more lively and realistic cobweb representation to take into account the difference in strength and length of threads. This model allows for visualization of highly unstructured dataset with large number of dimensions, common in Bigdata sources.	[ { "version": "v1", "created": "Mon, 24 Dec 2012 17:10:28 GMT" } ]	2013-01-03T00:00:00	[ [ "Prakash", "Aaditya", "" ] ]	TITLE: Reconstructing Self Organizing Maps as Spider Graphs for better visual interpretation of large unstructured datasets ABSTRACT: Self-Organizing Maps (SOM) are popular unsupervised artificial neural network used to reduce dimensions and visualize data. Visual interpretation from Self-Organizing Maps (SOM) has been limited due to grid approach of data representation, which makes inter-scenario analysis impossible. The paper proposes a new way to structure SOM. This model reconstructs SOM to show strength between variables as the threads of a cobweb and illuminate inter-scenario analysis. While Radar Graphs are very crude representation of spider web, this model uses more lively and realistic cobweb representation to take into account the difference in strength and length of threads. This model allows for visualization of highly unstructured dataset with large number of dimensions, common in Bigdata sources.
1205.4463	Salah A. Aly	Salah A. Aly	Pilgrims Face Recognition Dataset -- HUFRD	5 pages, 13 images, 1 table of a new HUFRD work	null	null	null	cs.CV cs.CY	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	In this work, we define a new pilgrims face recognition dataset, called HUFRD dataset. The new developed dataset presents various pilgrims' images taken from outside the Holy Masjid El-Harram in Makkah during the 2011-2012 Hajj and Umrah seasons. Such dataset will be used to test our developed facial recognition and detection algorithms, as well as assess in the missing and found recognition system \cite{crowdsensing}.	[ { "version": "v1", "created": "Sun, 20 May 2012 22:07:27 GMT" }, { "version": "v2", "created": "Sun, 30 Dec 2012 00:58:09 GMT" } ]	2013-01-01T00:00:00	[ [ "Aly", "Salah A.", "" ] ]	TITLE: Pilgrims Face Recognition Dataset -- HUFRD ABSTRACT: In this work, we define a new pilgrims face recognition dataset, called HUFRD dataset. The new developed dataset presents various pilgrims' images taken from outside the Holy Masjid El-Harram in Makkah during the 2011-2012 Hajj and Umrah seasons. Such dataset will be used to test our developed facial recognition and detection algorithms, as well as assess in the missing and found recognition system \cite{crowdsensing}.
1212.6659	Raphael Pelossof	Raphael Pelossof and Zhiliang Ying	Focus of Attention for Linear Predictors	9 pages, 4 figures. arXiv admin note: substantial text overlap with arXiv:1105.0382	null	null	null	stat.ML cs.AI cs.LG	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	We present a method to stop the evaluation of a prediction process when the result of the full evaluation is obvious. This trait is highly desirable in prediction tasks where a predictor evaluates all its features for every example in large datasets. We observe that some examples are easier to classify than others, a phenomenon which is characterized by the event when most of the features agree on the class of an example. By stopping the feature evaluation when encountering an easy- to-classify example, the predictor can achieve substantial gains in computation. Our method provides a natural attention mechanism for linear predictors where the predictor concentrates most of its computation on hard-to-classify examples and quickly discards easy-to-classify ones. By modifying a linear prediction algorithm such as an SVM or AdaBoost to include our attentive method we prove that the average number of features computed is O(sqrt(n log 1/sqrt(delta))) where n is the original number of features, and delta is the error rate incurred due to early stopping. We demonstrate the effectiveness of Attentive Prediction on MNIST, Real-sim, Gisette, and synthetic datasets.	[ { "version": "v1", "created": "Sat, 29 Dec 2012 20:23:48 GMT" } ]	2013-01-01T00:00:00	[ [ "Pelossof", "Raphael", "" ], [ "Ying", "Zhiliang", "" ] ]	TITLE: Focus of Attention for Linear Predictors ABSTRACT: We present a method to stop the evaluation of a prediction process when the result of the full evaluation is obvious. This trait is highly desirable in prediction tasks where a predictor evaluates all its features for every example in large datasets. We observe that some examples are easier to classify than others, a phenomenon which is characterized by the event when most of the features agree on the class of an example. By stopping the feature evaluation when encountering an easy- to-classify example, the predictor can achieve substantial gains in computation. Our method provides a natural attention mechanism for linear predictors where the predictor concentrates most of its computation on hard-to-classify examples and quickly discards easy-to-classify ones. By modifying a linear prediction algorithm such as an SVM or AdaBoost to include our attentive method we prove that the average number of features computed is O(sqrt(n log 1/sqrt(delta))) where n is the original number of features, and delta is the error rate incurred due to early stopping. We demonstrate the effectiveness of Attentive Prediction on MNIST, Real-sim, Gisette, and synthetic datasets.
1212.5637	Claudio Gentile	Nicolo' Cesa-Bianchi, Claudio Gentile, Fabio Vitale, Giovanni Zappella	Random Spanning Trees and the Prediction of Weighted Graphs	Appeared in ICML 2010	null	null	null	cs.LG stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	We investigate the problem of sequentially predicting the binary labels on the nodes of an arbitrary weighted graph. We show that, under a suitable parametrization of the problem, the optimal number of prediction mistakes can be characterized (up to logarithmic factors) by the cutsize of a random spanning tree of the graph. The cutsize is induced by the unknown adversarial labeling of the graph nodes. In deriving our characterization, we obtain a simple randomized algorithm achieving in expectation the optimal mistake bound on any polynomially connected weighted graph. Our algorithm draws a random spanning tree of the original graph and then predicts the nodes of this tree in constant expected amortized time and linear space. Experiments on real-world datasets show that our method compares well to both global (Perceptron) and local (label propagation) methods, while being generally faster in practice.	[ { "version": "v1", "created": "Fri, 21 Dec 2012 23:51:21 GMT" } ]	2012-12-27T00:00:00	[ [ "Cesa-Bianchi", "Nicolo'", "" ], [ "Gentile", "Claudio", "" ], [ "Vitale", "Fabio", "" ], [ "Zappella", "Giovanni", "" ] ]	TITLE: Random Spanning Trees and the Prediction of Weighted Graphs ABSTRACT: We investigate the problem of sequentially predicting the binary labels on the nodes of an arbitrary weighted graph. We show that, under a suitable parametrization of the problem, the optimal number of prediction mistakes can be characterized (up to logarithmic factors) by the cutsize of a random spanning tree of the graph. The cutsize is induced by the unknown adversarial labeling of the graph nodes. In deriving our characterization, we obtain a simple randomized algorithm achieving in expectation the optimal mistake bound on any polynomially connected weighted graph. Our algorithm draws a random spanning tree of the original graph and then predicts the nodes of this tree in constant expected amortized time and linear space. Experiments on real-world datasets show that our method compares well to both global (Perceptron) and local (label propagation) methods, while being generally faster in practice.
1212.5701	Matthew Zeiler	Matthew D. Zeiler	ADADELTA: An Adaptive Learning Rate Method	6 pages	null	null	null	cs.LG	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	We present a novel per-dimension learning rate method for gradient descent called ADADELTA. The method dynamically adapts over time using only first order information and has minimal computational overhead beyond vanilla stochastic gradient descent. The method requires no manual tuning of a learning rate and appears robust to noisy gradient information, different model architecture choices, various data modalities and selection of hyperparameters. We show promising results compared to other methods on the MNIST digit classification task using a single machine and on a large scale voice dataset in a distributed cluster environment.	[ { "version": "v1", "created": "Sat, 22 Dec 2012 15:46:49 GMT" } ]	2012-12-27T00:00:00	[ [ "Zeiler", "Matthew D.", "" ] ]	TITLE: ADADELTA: An Adaptive Learning Rate Method ABSTRACT: We present a novel per-dimension learning rate method for gradient descent called ADADELTA. The method dynamically adapts over time using only first order information and has minimal computational overhead beyond vanilla stochastic gradient descent. The method requires no manual tuning of a learning rate and appears robust to noisy gradient information, different model architecture choices, various data modalities and selection of hyperparameters. We show promising results compared to other methods on the MNIST digit classification task using a single machine and on a large scale voice dataset in a distributed cluster environment.
1212.6246	Radford M. Neal	Chunyi Wang and Radford M. Neal	Gaussian Process Regression with Heteroscedastic or Non-Gaussian Residuals	null	null	null	null	stat.ML cs.LG	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Gaussian Process (GP) regression models typically assume that residuals are Gaussian and have the same variance for all observations. However, applications with input-dependent noise (heteroscedastic residuals) frequently arise in practice, as do applications in which the residuals do not have a Gaussian distribution. In this paper, we propose a GP Regression model with a latent variable that serves as an additional unobserved covariate for the regression. This model (which we call GPLC) allows for heteroscedasticity since it allows the function to have a changing partial derivative with respect to this unobserved covariate. With a suitable covariance function, our GPLC model can handle (a) Gaussian residuals with input-dependent variance, or (b) non-Gaussian residuals with input-dependent variance, or (c) Gaussian residuals with constant variance. We compare our model, using synthetic datasets, with a model proposed by Goldberg, Williams and Bishop (1998), which we refer to as GPLV, which only deals with case (a), as well as a standard GP model which can handle only case (c). Markov Chain Monte Carlo methods are developed for both modelsl. Experiments show that when the data is heteroscedastic, both GPLC and GPLV give better results (smaller mean squared error and negative log-probability density) than standard GP regression. In addition, when the residual are Gaussian, our GPLC model is generally nearly as good as GPLV, while when the residuals are non-Gaussian, our GPLC model is better than GPLV.	[ { "version": "v1", "created": "Wed, 26 Dec 2012 20:45:48 GMT" } ]	2012-12-27T00:00:00	[ [ "Wang", "Chunyi", "" ], [ "Neal", "Radford M.", "" ] ]	TITLE: Gaussian Process Regression with Heteroscedastic or Non-Gaussian Residuals ABSTRACT: Gaussian Process (GP) regression models typically assume that residuals are Gaussian and have the same variance for all observations. However, applications with input-dependent noise (heteroscedastic residuals) frequently arise in practice, as do applications in which the residuals do not have a Gaussian distribution. In this paper, we propose a GP Regression model with a latent variable that serves as an additional unobserved covariate for the regression. This model (which we call GPLC) allows for heteroscedasticity since it allows the function to have a changing partial derivative with respect to this unobserved covariate. With a suitable covariance function, our GPLC model can handle (a) Gaussian residuals with input-dependent variance, or (b) non-Gaussian residuals with input-dependent variance, or (c) Gaussian residuals with constant variance. We compare our model, using synthetic datasets, with a model proposed by Goldberg, Williams and Bishop (1998), which we refer to as GPLV, which only deals with case (a), as well as a standard GP model which can handle only case (c). Markov Chain Monte Carlo methods are developed for both modelsl. Experiments show that when the data is heteroscedastic, both GPLC and GPLV give better results (smaller mean squared error and negative log-probability density) than standard GP regression. In addition, when the residual are Gaussian, our GPLC model is generally nearly as good as GPLV, while when the residuals are non-Gaussian, our GPLC model is better than GPLV.
1211.3089	Yuheng Hu	Yuheng Hu, Ajita John, Fei Wang, Subbarao Kambhampati	ET-LDA: Joint Topic Modeling for Aligning Events and their Twitter Feedback	reference error, delete for now	null	null	null	cs.SI cs.AI cs.CY	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	During broadcast events such as the Superbowl, the U.S. Presidential and Primary debates, etc., Twitter has become the de facto platform for crowds to share perspectives and commentaries about them. Given an event and an associated large-scale collection of tweets, there are two fundamental research problems that have been receiving increasing attention in recent years. One is to extract the topics covered by the event and the tweets; the other is to segment the event. So far these problems have been viewed separately and studied in isolation. In this work, we argue that these problems are in fact inter-dependent and should be addressed together. We develop a joint Bayesian model that performs topic modeling and event segmentation in one unified framework. We evaluate the proposed model both quantitatively and qualitatively on two large-scale tweet datasets associated with two events from different domains to show that it improves significantly over baseline models.	[ { "version": "v1", "created": "Tue, 13 Nov 2012 19:46:51 GMT" }, { "version": "v2", "created": "Fri, 21 Dec 2012 05:50:15 GMT" } ]	2012-12-24T00:00:00	[ [ "Hu", "Yuheng", "" ], [ "John", "Ajita", "" ], [ "Wang", "Fei", "" ], [ "Kambhampati", "Subbarao", "" ] ]	TITLE: ET-LDA: Joint Topic Modeling for Aligning Events and their Twitter Feedback ABSTRACT: During broadcast events such as the Superbowl, the U.S. Presidential and Primary debates, etc., Twitter has become the de facto platform for crowds to share perspectives and commentaries about them. Given an event and an associated large-scale collection of tweets, there are two fundamental research problems that have been receiving increasing attention in recent years. One is to extract the topics covered by the event and the tweets; the other is to segment the event. So far these problems have been viewed separately and studied in isolation. In this work, we argue that these problems are in fact inter-dependent and should be addressed together. We develop a joint Bayesian model that performs topic modeling and event segmentation in one unified framework. We evaluate the proposed model both quantitatively and qualitatively on two large-scale tweet datasets associated with two events from different domains to show that it improves significantly over baseline models.
1212.5265	Tamal Ghosh Tamal Ghosh	Tamal Ghosh, Pranab K Dan	An Effective Machine-Part Grouping Algorithm to Construct Manufacturing Cells	null	Proceedings of Conference on Industrial Engineering (NCIE 2011)	null	null	cs.CE	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	The machine-part cell formation problem consists of creating machine cells and their corresponding part families with the objective of minimizing the inter-cell and intra-cell movement while maximizing the machine utilization. This article demonstrates a hybrid clustering approach for the cell formation problem in cellular manufacturing that conjoins Sorenson s similarity coefficient based method to form the production cells. Computational results are shown over the test datasets obtained from the past literature. The hybrid technique is shown to outperform the other methods proposed in literature and including powerful soft computing approaches such as genetic algorithms, genetic programming by exceeding the solution quality on the test problems.	[ { "version": "v1", "created": "Thu, 20 Dec 2012 15:51:13 GMT" } ]	2012-12-24T00:00:00	[ [ "Ghosh", "Tamal", "" ], [ "Dan", "Pranab K", "" ] ]	TITLE: An Effective Machine-Part Grouping Algorithm to Construct Manufacturing Cells ABSTRACT: The machine-part cell formation problem consists of creating machine cells and their corresponding part families with the objective of minimizing the inter-cell and intra-cell movement while maximizing the machine utilization. This article demonstrates a hybrid clustering approach for the cell formation problem in cellular manufacturing that conjoins Sorenson s similarity coefficient based method to form the production cells. Computational results are shown over the test datasets obtained from the past literature. The hybrid technique is shown to outperform the other methods proposed in literature and including powerful soft computing approaches such as genetic algorithms, genetic programming by exceeding the solution quality on the test problems.
1212.4458	Dan Burger	Dan Burger, Keivan G. Stassun, Joshua Pepper, Robert J. Siverd, Martin A. Paegert, Nathan M. De Lee	Filtergraph: A Flexible Web Application for Instant Data Visualization of Astronomy Datasets	4 pages, 1 figure. Originally presented at the ADASS XXII Conference in Champaign, IL on November 6, 2012. Published in the conference proceedings by ASP Conference Series (revised to include URL of web application)	null	null	null	astro-ph.IM cs.SE	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Filtergraph is a web application being developed by the Vanderbilt Initiative in Data-intensive Astrophysics (VIDA) to flexibly handle a large variety of astronomy datasets. While current datasets at Vanderbilt are being used to search for eclipsing binaries and extrasolar planets, this system can be easily reconfigured for a wide variety of data sources. The user loads a flat-file dataset into Filtergraph which instantly generates an interactive data portal that can be easily shared with others. From this portal, the user can immediately generate scatter plots, histograms, and tables based on the dataset. Key features of the portal include the ability to filter the data in real time through user-specified criteria, the ability to select data by dragging on the screen, and the ability to perform arithmetic operations on the data in real time. The application is being optimized for speed in the context of very large datasets: for instance, plot generated from a stellar database of 3.1 million entries render in less than 2 seconds on a standard web server platform. This web application has been created using the Web2py web framework based on the Python programming language. Filtergraph is freely available at http://filtergraph.vanderbilt.edu/.	[ { "version": "v1", "created": "Tue, 18 Dec 2012 19:00:06 GMT" }, { "version": "v2", "created": "Wed, 19 Dec 2012 17:04:02 GMT" }, { "version": "v3", "created": "Thu, 20 Dec 2012 17:17:00 GMT" } ]	2012-12-21T00:00:00	[ [ "Burger", "Dan", "" ], [ "Stassun", "Keivan G.", "" ], [ "Pepper", "Joshua", "" ], [ "Siverd", "Robert J.", "" ], [ "Paegert", "Martin A.", "" ], [ "De Lee", "Nathan M.", "" ] ]	TITLE: Filtergraph: A Flexible Web Application for Instant Data Visualization of Astronomy Datasets ABSTRACT: Filtergraph is a web application being developed by the Vanderbilt Initiative in Data-intensive Astrophysics (VIDA) to flexibly handle a large variety of astronomy datasets. While current datasets at Vanderbilt are being used to search for eclipsing binaries and extrasolar planets, this system can be easily reconfigured for a wide variety of data sources. The user loads a flat-file dataset into Filtergraph which instantly generates an interactive data portal that can be easily shared with others. From this portal, the user can immediately generate scatter plots, histograms, and tables based on the dataset. Key features of the portal include the ability to filter the data in real time through user-specified criteria, the ability to select data by dragging on the screen, and the ability to perform arithmetic operations on the data in real time. The application is being optimized for speed in the context of very large datasets: for instance, plot generated from a stellar database of 3.1 million entries render in less than 2 seconds on a standard web server platform. This web application has been created using the Web2py web framework based on the Python programming language. Filtergraph is freely available at http://filtergraph.vanderbilt.edu/.
1212.4871	Ramin Norousi	Ramin Norousi, Stephan Wickles, Christoph Leidig, Thomas Becker, Volker J. Schmid, Roland Beckmann, Achim Tresch	Automatic post-picking using MAPPOS improves particle image detection from Cryo-EM micrographs	null	null	null	null	stat.ML cs.CV	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Cryo-electron microscopy (cryo-EM) studies using single particle reconstruction are extensively used to reveal structural information on macromolecular complexes. Aiming at the highest achievable resolution, state of the art electron microscopes automatically acquire thousands of high-quality micrographs. Particles are detected on and boxed out from each micrograph using fully- or semi-automated approaches. However, the obtained particles still require laborious manual post-picking classification, which is one major bottleneck for single particle analysis of large datasets. We introduce MAPPOS, a supervised post-picking strategy for the classification of boxed particle images, as additional strategy adding to the already efficient automated particle picking routines. MAPPOS employs machine learning techniques to train a robust classifier from a small number of characteristic image features. In order to accurately quantify the performance of MAPPOS we used simulated particle and non-particle images. In addition, we verified our method by applying it to an experimental cryo-EM dataset and comparing the results to the manual classification of the same dataset. Comparisons between MAPPOS and manual post-picking classification by several human experts demonstrated that merely a few hundred sample images are sufficient for MAPPOS to classify an entire dataset with a human-like performance. MAPPOS was shown to greatly accelerate the throughput of large datasets by reducing the manual workload by orders of magnitude while maintaining a reliable identification of non-particle images.	[ { "version": "v1", "created": "Wed, 19 Dec 2012 22:17:18 GMT" } ]	2012-12-21T00:00:00	[ [ "Norousi", "Ramin", "" ], [ "Wickles", "Stephan", "" ], [ "Leidig", "Christoph", "" ], [ "Becker", "Thomas", "" ], [ "Schmid", "Volker J.", "" ], [ "Beckmann", "Roland", "" ], [ "Tresch", "Achim", "" ] ]	TITLE: Automatic post-picking using MAPPOS improves particle image detection from Cryo-EM micrographs ABSTRACT: Cryo-electron microscopy (cryo-EM) studies using single particle reconstruction are extensively used to reveal structural information on macromolecular complexes. Aiming at the highest achievable resolution, state of the art electron microscopes automatically acquire thousands of high-quality micrographs. Particles are detected on and boxed out from each micrograph using fully- or semi-automated approaches. However, the obtained particles still require laborious manual post-picking classification, which is one major bottleneck for single particle analysis of large datasets. We introduce MAPPOS, a supervised post-picking strategy for the classification of boxed particle images, as additional strategy adding to the already efficient automated particle picking routines. MAPPOS employs machine learning techniques to train a robust classifier from a small number of characteristic image features. In order to accurately quantify the performance of MAPPOS we used simulated particle and non-particle images. In addition, we verified our method by applying it to an experimental cryo-EM dataset and comparing the results to the manual classification of the same dataset. Comparisons between MAPPOS and manual post-picking classification by several human experts demonstrated that merely a few hundred sample images are sufficient for MAPPOS to classify an entire dataset with a human-like performance. MAPPOS was shown to greatly accelerate the throughput of large datasets by reducing the manual workload by orders of magnitude while maintaining a reliable identification of non-particle images.
1212.4788	Dominik Grimm dg	Dominik Grimm, Bastian Greshake, Stefan Kleeberger, Christoph Lippert, Oliver Stegle, Bernhard Sch\"olkopf, Detlef Weigel and Karsten Borgwardt	easyGWAS: An integrated interspecies platform for performing genome-wide association studies	null	null	null	null	q-bio.GN cs.CE cs.DL stat.AP	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Motivation: The rapid growth in genome-wide association studies (GWAS) in plants and animals has brought about the need for a central resource that facilitates i) performing GWAS, ii) accessing data and results of other GWAS, and iii) enabling all users regardless of their background to exploit the latest statistical techniques without having to manage complex software and computing resources. Results: We present easyGWAS, a web platform that provides methods, tools and dynamic visualizations to perform and analyze GWAS. In addition, easyGWAS makes it simple to reproduce results of others, validate findings, and access larger sample sizes through merging of public datasets. Availability: Detailed method and data descriptions as well as tutorials are available in the supplementary materials. easyGWAS is available at http://easygwas.tuebingen.mpg.de/. Contact: [email protected]	[ { "version": "v1", "created": "Wed, 19 Dec 2012 18:39:06 GMT" } ]	2012-12-20T00:00:00	[ [ "Grimm", "Dominik", "" ], [ "Greshake", "Bastian", "" ], [ "Kleeberger", "Stefan", "" ], [ "Lippert", "Christoph", "" ], [ "Stegle", "Oliver", "" ], [ "Schölkopf", "Bernhard", "" ], [ "Weigel", "Detlef", "" ], [ "Borgwardt", "Karsten", "" ] ]	TITLE: easyGWAS: An integrated interspecies platform for performing genome-wide association studies ABSTRACT: Motivation: The rapid growth in genome-wide association studies (GWAS) in plants and animals has brought about the need for a central resource that facilitates i) performing GWAS, ii) accessing data and results of other GWAS, and iii) enabling all users regardless of their background to exploit the latest statistical techniques without having to manage complex software and computing resources. Results: We present easyGWAS, a web platform that provides methods, tools and dynamic visualizations to perform and analyze GWAS. In addition, easyGWAS makes it simple to reproduce results of others, validate findings, and access larger sample sizes through merging of public datasets. Availability: Detailed method and data descriptions as well as tutorials are available in the supplementary materials. easyGWAS is available at http://easygwas.tuebingen.mpg.de/. Contact: [email protected]
1212.4347	Bonggun Shin	Bonggun Shin, Alice Oh	Bayesian Group Nonnegative Matrix Factorization for EEG Analysis	null	null	null	null	cs.LG stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	We propose a generative model of a group EEG analysis, based on appropriate kernel assumptions on EEG data. We derive the variational inference update rule using various approximation techniques. The proposed model outperforms the current state-of-the-art algorithms in terms of common pattern extraction. The validity of the proposed model is tested on the BCI competition dataset.	[ { "version": "v1", "created": "Tue, 18 Dec 2012 13:35:38 GMT" } ]	2012-12-19T00:00:00	[ [ "Shin", "Bonggun", "" ], [ "Oh", "Alice", "" ] ]	TITLE: Bayesian Group Nonnegative Matrix Factorization for EEG Analysis ABSTRACT: We propose a generative model of a group EEG analysis, based on appropriate kernel assumptions on EEG data. We derive the variational inference update rule using various approximation techniques. The proposed model outperforms the current state-of-the-art algorithms in terms of common pattern extraction. The validity of the proposed model is tested on the BCI competition dataset.
1212.3938	Aurore Laurendeau	Aurore Laurendeau (ISTerre), Fabrice Cotton (ISTerre), Luis Fabian Bonilla	Nonstationary Stochastic Simulation of Strong Ground-Motion Time Histories : Application to the Japanese Database	10 pages; 15th World Conference on Earthquake Engineering, Lisbon : Portugal (2012)	null	null	null	stat.AP physics.geo-ph	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	For earthquake-resistant design, engineering seismologists employ time-history analysis for nonlinear simulations. The nonstationary stochastic method previously developed by Pousse et al. (2006) has been updated. This method has the advantage of being both simple, fast and taking into account the basic concepts of seismology (Brune's source, realistic time envelope function, nonstationarity and ground-motion variability). Time-domain simulations are derived from the signal spectrogram and depend on few ground-motion parameters: Arias intensity, significant relative duration and central frequency. These indicators are obtained from empirical attenuation equations that relate them to the magnitude of the event, the source-receiver distance, and the site conditions. We improve the nonstationary stochastic method by using new functional forms (new surface rock dataset, analysis of both intra-event and inter-event residuals, consideration of the scaling relations and VS30), by assessing the central frequency with S-transform and by better considering the stress drop variability.	[ { "version": "v1", "created": "Mon, 17 Dec 2012 08:47:29 GMT" } ]	2012-12-18T00:00:00	[ [ "Laurendeau", "Aurore", "", "ISTerre" ], [ "Cotton", "Fabrice", "", "ISTerre" ], [ "Bonilla", "Luis Fabian", "" ] ]	TITLE: Nonstationary Stochastic Simulation of Strong Ground-Motion Time Histories : Application to the Japanese Database ABSTRACT: For earthquake-resistant design, engineering seismologists employ time-history analysis for nonlinear simulations. The nonstationary stochastic method previously developed by Pousse et al. (2006) has been updated. This method has the advantage of being both simple, fast and taking into account the basic concepts of seismology (Brune's source, realistic time envelope function, nonstationarity and ground-motion variability). Time-domain simulations are derived from the signal spectrogram and depend on few ground-motion parameters: Arias intensity, significant relative duration and central frequency. These indicators are obtained from empirical attenuation equations that relate them to the magnitude of the event, the source-receiver distance, and the site conditions. We improve the nonstationary stochastic method by using new functional forms (new surface rock dataset, analysis of both intra-event and inter-event residuals, consideration of the scaling relations and VS30), by assessing the central frequency with S-transform and by better considering the stress drop variability.
1212.3964	Sourav Dutta	Suman K. Bera, Sourav Dutta, Ankur Narang and Souvik Bhattacherjee	Advanced Bloom Filter Based Algorithms for Efficient Approximate Data De-Duplication in Streams	41 pages	null	null	null	cs.IR	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Applications involving telecommunication call data records, web pages, online transactions, medical records, stock markets, climate warning systems, etc., necessitate efficient management and processing of such massively exponential amount of data from diverse sources. De-duplication or Intelligent Compression in streaming scenarios for approximate identification and elimination of duplicates from such unbounded data stream is a greater challenge given the real-time nature of data arrival. Stable Bloom Filters (SBF) addresses this problem to a certain extent. . In this work, we present several novel algorithms for the problem of approximate detection of duplicates in data streams. We propose the Reservoir Sampling based Bloom Filter (RSBF) combining the working principle of reservoir sampling and Bloom Filters. We also present variants of the novel Biased Sampling based Bloom Filter (BSBF) based on biased sampling concepts. We also propose a randomized load balanced variant of the sampling Bloom Filter approach to efficiently tackle the duplicate detection. In this work, we thus provide a generic framework for de-duplication using Bloom Filters. Using detailed theoretical analysis we prove analytical bounds on the false positive rate, false negative rate and convergence rate of the proposed structures. We exhibit that our models clearly outperform the existing methods. We also demonstrate empirical analysis of the structures using real-world datasets (3 million records) and also with synthetic datasets (1 billion records) capturing various input distributions.	[ { "version": "v1", "created": "Mon, 17 Dec 2012 11:47:09 GMT" } ]	2012-12-18T00:00:00	[ [ "Bera", "Suman K.", "" ], [ "Dutta", "Sourav", "" ], [ "Narang", "Ankur", "" ], [ "Bhattacherjee", "Souvik", "" ] ]	TITLE: Advanced Bloom Filter Based Algorithms for Efficient Approximate Data De-Duplication in Streams ABSTRACT: Applications involving telecommunication call data records, web pages, online transactions, medical records, stock markets, climate warning systems, etc., necessitate efficient management and processing of such massively exponential amount of data from diverse sources. De-duplication or Intelligent Compression in streaming scenarios for approximate identification and elimination of duplicates from such unbounded data stream is a greater challenge given the real-time nature of data arrival. Stable Bloom Filters (SBF) addresses this problem to a certain extent. . In this work, we present several novel algorithms for the problem of approximate detection of duplicates in data streams. We propose the Reservoir Sampling based Bloom Filter (RSBF) combining the working principle of reservoir sampling and Bloom Filters. We also present variants of the novel Biased Sampling based Bloom Filter (BSBF) based on biased sampling concepts. We also propose a randomized load balanced variant of the sampling Bloom Filter approach to efficiently tackle the duplicate detection. In this work, we thus provide a generic framework for de-duplication using Bloom Filters. Using detailed theoretical analysis we prove analytical bounds on the false positive rate, false negative rate and convergence rate of the proposed structures. We exhibit that our models clearly outperform the existing methods. We also demonstrate empirical analysis of the structures using real-world datasets (3 million records) and also with synthetic datasets (1 billion records) capturing various input distributions.
1212.3390	A Majumder	Anirban Majumder and Nisheeth Shrivastava	Know Your Personalization: Learning Topic level Personalization in Online Services	privacy, personalization	null	null	null	cs.LG cs.IR	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Online service platforms (OSPs), such as search engines, news-websites, ad-providers, etc., serve highly pe rsonalized content to the user, based on the profile extracted from his history with the OSP. Although personalization (generally) leads to a better user experience, it also raises privacy concerns for the user---he does not know what is present in his profile and more importantly, what is being used to per sonalize content for him. In this paper, we capture OSP's personalization for an user in a new data structure called the person alization vector ($\eta$), which is a weighted vector over a set of topics, and present techniques to compute it for users of an OSP. Our approach treats OSPs as black-boxes, and extracts $\eta$ by mining only their output, specifical ly, the personalized (for an user) and vanilla (without any user information) contents served, and the differences in these content. We formulate a new model called Latent Topic Personalization (LTP) that captures the personalization vector into a learning framework and present efficient inference algorithms for it. We do extensive experiments for search result personalization using both data from real Google users and synthetic datasets. Our results show high accuracy (R-pre = 84%) of LTP in finding personalized topics. For Google data, our qualitative results show how LTP can also identifies evidences---queries for results on a topic with high $\eta$ value were re-ranked. Finally, we show how our approach can be used to build a new Privacy evaluation framework focused at end-user privacy on commercial OSPs.	[ { "version": "v1", "created": "Fri, 14 Dec 2012 04:12:21 GMT" } ]	2012-12-17T00:00:00	[ [ "Majumder", "Anirban", "" ], [ "Shrivastava", "Nisheeth", "" ] ]	TITLE: Know Your Personalization: Learning Topic level Personalization in Online Services ABSTRACT: Online service platforms (OSPs), such as search engines, news-websites, ad-providers, etc., serve highly pe rsonalized content to the user, based on the profile extracted from his history with the OSP. Although personalization (generally) leads to a better user experience, it also raises privacy concerns for the user---he does not know what is present in his profile and more importantly, what is being used to per sonalize content for him. In this paper, we capture OSP's personalization for an user in a new data structure called the person alization vector ($\eta$), which is a weighted vector over a set of topics, and present techniques to compute it for users of an OSP. Our approach treats OSPs as black-boxes, and extracts $\eta$ by mining only their output, specifical ly, the personalized (for an user) and vanilla (without any user information) contents served, and the differences in these content. We formulate a new model called Latent Topic Personalization (LTP) that captures the personalization vector into a learning framework and present efficient inference algorithms for it. We do extensive experiments for search result personalization using both data from real Google users and synthetic datasets. Our results show high accuracy (R-pre = 84%) of LTP in finding personalized topics. For Google data, our qualitative results show how LTP can also identifies evidences---queries for results on a topic with high $\eta$ value were re-ranked. Finally, we show how our approach can be used to build a new Privacy evaluation framework focused at end-user privacy on commercial OSPs.
1212.2981	Celine Beauval	C\'eline Beauval (ISTerre), Hilal Tasan (ISTerre), Aurore Laurendeau (ISTerre), Elise Delavaud, Fabrice Cotton (ISTerre), Philippe Gu\'eguen (ISTerre), Nicolas Kuehn	On the Testing of Ground--Motion Prediction Equations against Small--Magnitude Data	null	Bulletin of the Seismological Society of America 102, 5 (2012) 1994-2007	10.1785/0120110271	null	physics.geo-ph	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Ground-motion prediction equations (GMPE) are essential in probabilistic seismic hazard studies for estimating the ground motions generated by the seismic sources. In low seismicity regions, only weak motions are available in the lifetime of accelerometric networks, and the equations selected for the probabilistic studies are usually models established from foreign data. Although most ground-motion prediction equations have been developed for magnitudes 5 and above, the minimum magnitude often used in probabilistic studies in low seismicity regions is smaller. Desaggregations have shown that, at return periods of engineering interest, magnitudes lower than 5 can be contributing to the hazard. This paper presents the testing of several GMPEs selected in current international and national probabilistic projects against weak motions recorded in France (191 recordings with source-site distances up to 300km, 3.8\leqMw\leq4.5). The method is based on the loglikelihood value proposed by Scherbaum et al. (2009). The best fitting models (approximately 2.5\leqLLH\leq3.5) over the whole frequency range are the Cauzzi and Faccioli (2008), Akkar and Bommer (2010) and Abrahamson and Silva (2008) models. No significant regional variation of ground motions is highlighted, and the magnitude scaling could be predominant in the control of ground-motion amplitudes. Furthermore, we take advantage of a rich Japanese dataset to run tests on randomly selected low-magnitude subsets, and check that a dataset of ~190 observations, same size as the French dataset, is large enough to obtain stable LLH estimates. Additionally we perform the tests against larger magnitudes (5-7) from the Japanese dataset. The ranking of models is partially modified, indicating a magnitude scaling effect for some of the models, and showing that extrapolating testing results obtained from low magnitude ranges to higher magnitude ranges is not straightforward.	[ { "version": "v1", "created": "Wed, 12 Dec 2012 21:01:39 GMT" } ]	2012-12-14T00:00:00	[ [ "Beauval", "Céline", "", "ISTerre" ], [ "Tasan", "Hilal", "", "ISTerre" ], [ "Laurendeau", "Aurore", "", "ISTerre" ], [ "Delavaud", "Elise", "", "ISTerre" ], [ "Cotton", "Fabrice", "", "ISTerre" ], [ "Guéguen", "Philippe", "", "ISTerre" ], [ "Kuehn", "Nicolas", "" ] ]	TITLE: On the Testing of Ground--Motion Prediction Equations against Small--Magnitude Data ABSTRACT: Ground-motion prediction equations (GMPE) are essential in probabilistic seismic hazard studies for estimating the ground motions generated by the seismic sources. In low seismicity regions, only weak motions are available in the lifetime of accelerometric networks, and the equations selected for the probabilistic studies are usually models established from foreign data. Although most ground-motion prediction equations have been developed for magnitudes 5 and above, the minimum magnitude often used in probabilistic studies in low seismicity regions is smaller. Desaggregations have shown that, at return periods of engineering interest, magnitudes lower than 5 can be contributing to the hazard. This paper presents the testing of several GMPEs selected in current international and national probabilistic projects against weak motions recorded in France (191 recordings with source-site distances up to 300km, 3.8\leqMw\leq4.5). The method is based on the loglikelihood value proposed by Scherbaum et al. (2009). The best fitting models (approximately 2.5\leqLLH\leq3.5) over the whole frequency range are the Cauzzi and Faccioli (2008), Akkar and Bommer (2010) and Abrahamson and Silva (2008) models. No significant regional variation of ground motions is highlighted, and the magnitude scaling could be predominant in the control of ground-motion amplitudes. Furthermore, we take advantage of a rich Japanese dataset to run tests on randomly selected low-magnitude subsets, and check that a dataset of ~190 observations, same size as the French dataset, is large enough to obtain stable LLH estimates. Additionally we perform the tests against larger magnitudes (5-7) from the Japanese dataset. The ranking of models is partially modified, indicating a magnitude scaling effect for some of the models, and showing that extrapolating testing results obtained from low magnitude ranges to higher magnitude ranges is not straightforward.
1212.3013	Gabriele Modena	K. Massoudi, G. Modena	Product/Brand extraction from WikiPedia	17 pages. Manuscript first creation date: November 27, 2009. At the time of first creation both authors were affiliated with the University of Amsterdam (The Netherlands)	null	null	null	cs.IR cs.AI	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	In this paper we describe the task of extracting product and brand pages from wikipedia. We present an experimental environment and setup built on top of a dataset of wikipedia pages we collected. We introduce a method for recognition of product pages modelled as a boolean probabilistic classification task. We show that this approach can lead to promising results and we discuss alternative approaches we considered.	[ { "version": "v1", "created": "Wed, 12 Dec 2012 23:25:46 GMT" } ]	2012-12-14T00:00:00	[ [ "Massoudi", "K.", "" ], [ "Modena", "G.", "" ] ]	TITLE: Product/Brand extraction from WikiPedia ABSTRACT: In this paper we describe the task of extracting product and brand pages from wikipedia. We present an experimental environment and setup built on top of a dataset of wikipedia pages we collected. We introduce a method for recognition of product pages modelled as a boolean probabilistic classification task. We show that this approach can lead to promising results and we discuss alternative approaches we considered.
1212.3152	Benjamin Laken	Benjamin A. Laken, Jasa \v{C}alogovi\'c, Tariq Shahbaz and Enric Pall\'e	Examining a solar climate link in diurnal temperature ranges	18 pages, 7 figures, 1 table	Laken B.A., J. Calogovic, T. Shahbaz, & E. Palle (2012) Examining a solar-climate link in diurnal temperature ranges. Journal of Geophysical Research, 117, D18112, 9PP	10.1029/2012JD17683	null	physics.ao-ph	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	A recent study has suggested a link between the surface level diurnal temperature range (DTR) and variations in the cosmic ray (CR) flux. As the DTR is an effective proxy for cloud cover, this result supports the notion that widespread cloud changes may be induced by the CR flux. If confirmed, this would have significant implications for our understanding of natural climate forcings. Here, we perform a detailed investigation of the relationships between DTR and solar activity (total solar irradiance and the CR flux) from more than 60 years of NCEP/NCAR reanalysis data and observations from meteorological station data. We find no statistically significant evidence to suggest that the DTR is connected to either long-term solar periodicities (11 or 1.68 year) or short-term (daily-timescale) fluctuations in solar activity, and we attribute previous reports on the contrary to an incorrect estimation of the statistical significance of the data. If a CR-DTR relationship exists, based on the estimated noise in DTR composites during Forbush decrease (FD) events, the DTR response would need to be larger than 0.03{\deg}C per 1% increase in the CR flux to be reliably detected. Compared with a much smaller rough estimate of -0.005{\deg}C per 1% increase in the CR flux expected if previous claims that FD events cause reductions in the cloud cover are valid, we conclude it is not possible to detect a solar related responses in station-based or reanalysis-based DTR datasets related to a hypothesized CR-cloud link, as potential signals would be drowned in noise.	[ { "version": "v1", "created": "Thu, 13 Dec 2012 12:42:43 GMT" } ]	2012-12-14T00:00:00	[ [ "Laken", "Benjamin A.", "" ], [ "Čalogović", "Jasa", "" ], [ "Shahbaz", "Tariq", "" ], [ "Pallé", "Enric", "" ] ]	TITLE: Examining a solar climate link in diurnal temperature ranges ABSTRACT: A recent study has suggested a link between the surface level diurnal temperature range (DTR) and variations in the cosmic ray (CR) flux. As the DTR is an effective proxy for cloud cover, this result supports the notion that widespread cloud changes may be induced by the CR flux. If confirmed, this would have significant implications for our understanding of natural climate forcings. Here, we perform a detailed investigation of the relationships between DTR and solar activity (total solar irradiance and the CR flux) from more than 60 years of NCEP/NCAR reanalysis data and observations from meteorological station data. We find no statistically significant evidence to suggest that the DTR is connected to either long-term solar periodicities (11 or 1.68 year) or short-term (daily-timescale) fluctuations in solar activity, and we attribute previous reports on the contrary to an incorrect estimation of the statistical significance of the data. If a CR-DTR relationship exists, based on the estimated noise in DTR composites during Forbush decrease (FD) events, the DTR response would need to be larger than 0.03{\deg}C per 1% increase in the CR flux to be reliably detected. Compared with a much smaller rough estimate of -0.005{\deg}C per 1% increase in the CR flux expected if previous claims that FD events cause reductions in the cloud cover are valid, we conclude it is not possible to detect a solar related responses in station-based or reanalysis-based DTR datasets related to a hypothesized CR-cloud link, as potential signals would be drowned in noise.
1212.3287	Celine Beauval	C\'eline Beauval (ISTerre), F. Cotton (ISTerre), N. Abrahamson, N. Theodulidis (ITSAK), E. Delavaud (ISTerre), L. Rodriguez (ISTerre), F. Scherbaum, A. Haendel	Regional differences in subduction ground motions	10 pages	World Conference on Earthquake Engineering, Lisbonne : Portugal (2012)	null	null	physics.geo-ph	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	A few ground-motion prediction models have been published in the last years, for predicting ground motions produced by interface and intraslab earthquakes. When one must carry out a probabilistic seismic hazard analysis in a region including a subduction zone, GMPEs must be selected to feed a logic tree. In the present study, the aim is to identify which models provide the best fit to the dataset M6+, global or local models. The subduction regions considered are Japan, Taiwan, Central and South America, and Greece. Most of the data comes from the database built to develop the new BCHydro subduction global GMPE (Abrahamson et al., submitted). We show that this model is among best-fitting models in all cases, followed closely by Zhao et al. (2006), whereas the local Lin and Lee (2008) is well predicting the data in Taiwan and also in Greece. The Scherbaum et al. (2009) LLH method prove to be efficient in providing one number quantifying the overall fit, but additional analysis on the between-event and within-event variabilities are mandatory, to control if median prediction per event and/or variability within an event is within the scatter predicted by the model.	[ { "version": "v1", "created": "Thu, 13 Dec 2012 19:41:24 GMT" } ]	2012-12-14T00:00:00	[ [ "Beauval", "Céline", "", "ISTerre" ], [ "Cotton", "F.", "", "ISTerre" ], [ "Abrahamson", "N.", "", "ITSAK" ], [ "Theodulidis", "N.", "", "ITSAK" ], [ "Delavaud", "E.", "", "ISTerre" ], [ "Rodriguez", "L.", "", "ISTerre" ], [ "Scherbaum", "F.", "" ], [ "Haendel", "A.", "" ] ]	TITLE: Regional differences in subduction ground motions ABSTRACT: A few ground-motion prediction models have been published in the last years, for predicting ground motions produced by interface and intraslab earthquakes. When one must carry out a probabilistic seismic hazard analysis in a region including a subduction zone, GMPEs must be selected to feed a logic tree. In the present study, the aim is to identify which models provide the best fit to the dataset M6+, global or local models. The subduction regions considered are Japan, Taiwan, Central and South America, and Greece. Most of the data comes from the database built to develop the new BCHydro subduction global GMPE (Abrahamson et al., submitted). We show that this model is among best-fitting models in all cases, followed closely by Zhao et al. (2006), whereas the local Lin and Lee (2008) is well predicting the data in Taiwan and also in Greece. The Scherbaum et al. (2009) LLH method prove to be efficient in providing one number quantifying the overall fit, but additional analysis on the between-event and within-event variabilities are mandatory, to control if median prediction per event and/or variability within an event is within the scatter predicted by the model.
1212.2692	Ghazali Osman	Ghazali Osman, Muhammad Suzuri Hitam and Mohd Nasir Ismail	Enhanced skin colour classifier using RGB Ratio model	14 pages; International Journal on Soft Computing (IJSC) Vol.3, No.4, November 2012	null	null	null	cs.CV	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Skin colour detection is frequently been used for searching people, face detection, pornographic filtering and hand tracking. The presence of skin or non-skin in digital image can be determined by manipulating pixels colour or pixels texture. The main problem in skin colour detection is to represent the skin colour distribution model that is invariant or least sensitive to changes in illumination condition. Another problem comes from the fact that many objects in the real world may possess almost similar skin-tone colour such as wood, leather, skin-coloured clothing, hair and sand. Moreover, skin colour is different between races and can be different from a person to another, even with people of the same ethnicity. Finally, skin colour will appear a little different when different types of camera are used to capture the object or scene. The objective in this study is to develop a skin colour classifier based on pixel-based using RGB ratio model. The RGB ratio model is a newly proposed method that belongs under the category of an explicitly defined skin region model. This skin classifier was tested with SIdb dataset and two benchmark datasets; UChile and TDSD datasets to measure classifier performance. The performance of skin classifier was measured based on true positive (TF) and false positive (FP) indicator. This newly proposed model was compared with Kovac, Saleh and Swift models. The experimental results showed that the RGB ratio model outperformed all the other models in term of detection rate. The RGB ratio model is able to reduce FP detection that caused by reddish objects colour as well as be able to detect darkened skin and skin covered by shadow.	[ { "version": "v1", "created": "Wed, 12 Dec 2012 03:01:00 GMT" } ]	2012-12-13T00:00:00	[ [ "Osman", "Ghazali", "" ], [ "Hitam", "Muhammad Suzuri", "" ], [ "Ismail", "Mohd Nasir", "" ] ]	TITLE: Enhanced skin colour classifier using RGB Ratio model ABSTRACT: Skin colour detection is frequently been used for searching people, face detection, pornographic filtering and hand tracking. The presence of skin or non-skin in digital image can be determined by manipulating pixels colour or pixels texture. The main problem in skin colour detection is to represent the skin colour distribution model that is invariant or least sensitive to changes in illumination condition. Another problem comes from the fact that many objects in the real world may possess almost similar skin-tone colour such as wood, leather, skin-coloured clothing, hair and sand. Moreover, skin colour is different between races and can be different from a person to another, even with people of the same ethnicity. Finally, skin colour will appear a little different when different types of camera are used to capture the object or scene. The objective in this study is to develop a skin colour classifier based on pixel-based using RGB ratio model. The RGB ratio model is a newly proposed method that belongs under the category of an explicitly defined skin region model. This skin classifier was tested with SIdb dataset and two benchmark datasets; UChile and TDSD datasets to measure classifier performance. The performance of skin classifier was measured based on true positive (TF) and false positive (FP) indicator. This newly proposed model was compared with Kovac, Saleh and Swift models. The experimental results showed that the RGB ratio model outperformed all the other models in term of detection rate. The RGB ratio model is able to reduce FP detection that caused by reddish objects colour as well as be able to detect darkened skin and skin covered by shadow.
1212.2823	Shuran Song	Shuran Song, Jianxiong Xiao	Tracking Revisited using RGBD Camera: Baseline and Benchmark	null	null	null	null	cs.CV	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Although there has been significant progress in the past decade,tracking is still a very challenging computer vision task, due to problems such as occlusion and model drift.Recently, the increased popularity of depth sensors e.g. Microsoft Kinect has made it easy to obtain depth data at low cost.This may be a game changer for tracking, since depth information can be used to prevent model drift and handle occlusion.In this paper, we construct a benchmark dataset of 100 RGBD videos with high diversity, including deformable objects, various occlusion conditions and moving cameras. We propose a very simple but strong baseline model for RGBD tracking, and present a quantitative comparison of several state-of-the-art tracking algorithms.Experimental results show that including depth information and reasoning about occlusion significantly improves tracking performance. The datasets, evaluation details, source code for the baseline algorithm, and instructions for submitting new models will be made available online after acceptance.	[ { "version": "v1", "created": "Wed, 12 Dec 2012 14:02:41 GMT" } ]	2012-12-13T00:00:00	[ [ "Song", "Shuran", "" ], [ "Xiao", "Jianxiong", "" ] ]	TITLE: Tracking Revisited using RGBD Camera: Baseline and Benchmark ABSTRACT: Although there has been significant progress in the past decade,tracking is still a very challenging computer vision task, due to problems such as occlusion and model drift.Recently, the increased popularity of depth sensors e.g. Microsoft Kinect has made it easy to obtain depth data at low cost.This may be a game changer for tracking, since depth information can be used to prevent model drift and handle occlusion.In this paper, we construct a benchmark dataset of 100 RGBD videos with high diversity, including deformable objects, various occlusion conditions and moving cameras. We propose a very simple but strong baseline model for RGBD tracking, and present a quantitative comparison of several state-of-the-art tracking algorithms.Experimental results show that including depth information and reasoning about occlusion significantly improves tracking performance. The datasets, evaluation details, source code for the baseline algorithm, and instructions for submitting new models will be made available online after acceptance.
1212.2468	David Maxwell Chickering	David Maxwell Chickering, Christopher Meek, David Heckerman	Large-Sample Learning of Bayesian Networks is NP-Hard	Appears in Proceedings of the Nineteenth Conference on Uncertainty in Artificial Intelligence (UAI2003)	null	null	UAI-P-2003-PG-124-133	cs.LG cs.AI stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	In this paper, we provide new complexity results for algorithms that learn discrete-variable Bayesian networks from data. Our results apply whenever the learning algorithm uses a scoring criterion that favors the simplest model able to represent the generative distribution exactly. Our results therefore hold whenever the learning algorithm uses a consistent scoring criterion and is applied to a sufficiently large dataset. We show that identifying high-scoring structures is hard, even when we are given an independence oracle, an inference oracle, and/or an information oracle. Our negative results also apply to the learning of discrete-variable Bayesian networks in which each node has at most k parents, for all k > 3.	[ { "version": "v1", "created": "Fri, 19 Oct 2012 15:04:28 GMT" } ]	2012-12-12T00:00:00	[ [ "Chickering", "David Maxwell", "" ], [ "Meek", "Christopher", "" ], [ "Heckerman", "David", "" ] ]	TITLE: Large-Sample Learning of Bayesian Networks is NP-Hard ABSTRACT: In this paper, we provide new complexity results for algorithms that learn discrete-variable Bayesian networks from data. Our results apply whenever the learning algorithm uses a scoring criterion that favors the simplest model able to represent the generative distribution exactly. Our results therefore hold whenever the learning algorithm uses a consistent scoring criterion and is applied to a sufficiently large dataset. We show that identifying high-scoring structures is hard, even when we are given an independence oracle, an inference oracle, and/or an information oracle. Our negative results also apply to the learning of discrete-variable Bayesian networks in which each node has at most k parents, for all k > 3.
1212.2478	Rong Jin	Rong Jin, Luo Si, ChengXiang Zhai	Preference-based Graphic Models for Collaborative Filtering	Appears in Proceedings of the Nineteenth Conference on Uncertainty in Artificial Intelligence (UAI2003)	null	null	UAI-P-2003-PG-329-336	cs.IR	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Collaborative filtering is a very useful general technique for exploiting the preference patterns of a group of users to predict the utility of items to a particular user. Previous research has studied several probabilistic graphic models for collaborative filtering with promising results. However, while these models have succeeded in capturing the similarity among users and items in one way or the other, none of them has considered the fact that users with similar interests in items can have very different rating patterns; some users tend to assign a higher rating to all items than other users. In this paper, we propose and study of two new graphic models that address the distinction between user preferences and ratings. In one model, called the decoupled model, we introduce two different variables to decouple a users preferences FROM his ratings. IN the other, called the preference model, we model the orderings OF items preferred BY a USER, rather than the USERs numerical ratings of items. Empirical study over two datasets of movie ratings shows that appropriate modeling of the distinction between user preferences and ratings improves the performance substantially and consistently. Specifically, the proposed decoupled model outperforms all five existing approaches that we compare with significantly, but the preference model is not very successful. These results suggest that explicit modeling of the underlying user preferences is very important for collaborative filtering, but we can not afford ignoring the rating information completely.	[ { "version": "v1", "created": "Fri, 19 Oct 2012 15:06:09 GMT" } ]	2012-12-12T00:00:00	[ [ "Jin", "Rong", "" ], [ "Si", "Luo", "" ], [ "Zhai", "ChengXiang", "" ] ]	TITLE: Preference-based Graphic Models for Collaborative Filtering ABSTRACT: Collaborative filtering is a very useful general technique for exploiting the preference patterns of a group of users to predict the utility of items to a particular user. Previous research has studied several probabilistic graphic models for collaborative filtering with promising results. However, while these models have succeeded in capturing the similarity among users and items in one way or the other, none of them has considered the fact that users with similar interests in items can have very different rating patterns; some users tend to assign a higher rating to all items than other users. In this paper, we propose and study of two new graphic models that address the distinction between user preferences and ratings. In one model, called the decoupled model, we introduce two different variables to decouple a users preferences FROM his ratings. IN the other, called the preference model, we model the orderings OF items preferred BY a USER, rather than the USERs numerical ratings of items. Empirical study over two datasets of movie ratings shows that appropriate modeling of the distinction between user preferences and ratings improves the performance substantially and consistently. Specifically, the proposed decoupled model outperforms all five existing approaches that we compare with significantly, but the preference model is not very successful. These results suggest that explicit modeling of the underlying user preferences is very important for collaborative filtering, but we can not afford ignoring the rating information completely.
1212.2483	Amir Globerson	Amir Globerson, Gal Chechik, Naftali Tishby	Sufficient Dimensionality Reduction with Irrelevant Statistics	Appears in Proceedings of the Nineteenth Conference on Uncertainty in Artificial Intelligence (UAI2003)	null	null	UAI-P-2003-PG-281-288	cs.LG stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	The problem of finding a reduced dimensionality representation of categorical variables while preserving their most relevant characteristics is fundamental for the analysis of complex data. Specifically, given a co-occurrence matrix of two variables, one often seeks a compact representation of one variable which preserves information about the other variable. We have recently introduced ``Sufficient Dimensionality Reduction' [GT-2003], a method that extracts continuous reduced dimensional features whose measurements (i.e., expectation values) capture maximal mutual information among the variables. However, such measurements often capture information that is irrelevant for a given task. Widely known examples are illumination conditions, which are irrelevant as features for face recognition, writing style which is irrelevant as a feature for content classification, and intonation which is irrelevant as a feature for speech recognition. Such irrelevance cannot be deduced apriori, since it depends on the details of the task, and is thus inherently ill defined in the purely unsupervised case. Separating relevant from irrelevant features can be achieved using additional side data that contains such irrelevant structures. This approach was taken in [CT-2002], extending the information bottleneck method, which uses clustering to compress the data. Here we use this side-information framework to identify features whose measurements are maximally informative for the original data set, but carry as little information as possible on a side data set. In statistical terms this can be understood as extracting statistics which are maximally sufficient for the original dataset, while simultaneously maximally ancillary for the side dataset. We formulate this tradeoff as a constrained optimization problem and characterize its solutions. We then derive a gradient descent algorithm for this problem, which is based on the Generalized Iterative Scaling method for finding maximum entropy distributions. The method is demonstrated on synthetic data, as well as on real face recognition datasets, and is shown to outperform standard methods such as oriented PCA.	[ { "version": "v1", "created": "Fri, 19 Oct 2012 15:05:46 GMT" } ]	2012-12-12T00:00:00	[ [ "Globerson", "Amir", "" ], [ "Chechik", "Gal", "" ], [ "Tishby", "Naftali", "" ] ]	TITLE: Sufficient Dimensionality Reduction with Irrelevant Statistics ABSTRACT: The problem of finding a reduced dimensionality representation of categorical variables while preserving their most relevant characteristics is fundamental for the analysis of complex data. Specifically, given a co-occurrence matrix of two variables, one often seeks a compact representation of one variable which preserves information about the other variable. We have recently introduced ``Sufficient Dimensionality Reduction' [GT-2003], a method that extracts continuous reduced dimensional features whose measurements (i.e., expectation values) capture maximal mutual information among the variables. However, such measurements often capture information that is irrelevant for a given task. Widely known examples are illumination conditions, which are irrelevant as features for face recognition, writing style which is irrelevant as a feature for content classification, and intonation which is irrelevant as a feature for speech recognition. Such irrelevance cannot be deduced apriori, since it depends on the details of the task, and is thus inherently ill defined in the purely unsupervised case. Separating relevant from irrelevant features can be achieved using additional side data that contains such irrelevant structures. This approach was taken in [CT-2002], extending the information bottleneck method, which uses clustering to compress the data. Here we use this side-information framework to identify features whose measurements are maximally informative for the original data set, but carry as little information as possible on a side data set. In statistical terms this can be understood as extracting statistics which are maximally sufficient for the original dataset, while simultaneously maximally ancillary for the side dataset. We formulate this tradeoff as a constrained optimization problem and characterize its solutions. We then derive a gradient descent algorithm for this problem, which is based on the Generalized Iterative Scaling method for finding maximum entropy distributions. The method is demonstrated on synthetic data, as well as on real face recognition datasets, and is shown to outperform standard methods such as oriented PCA.
1212.2546	Jonathan Masci	Jonathan Masci and Jes\'us Angulo and J\"urgen Schmidhuber	A Learning Framework for Morphological Operators using Counter-Harmonic Mean	Submitted to ISMM'13	null	null	null	cs.CV	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	We present a novel framework for learning morphological operators using counter-harmonic mean. It combines concepts from morphology and convolutional neural networks. A thorough experimental validation analyzes basic morphological operators dilation and erosion, opening and closing, as well as the much more complex top-hat transform, for which we report a real-world application from the steel industry. Using online learning and stochastic gradient descent, our system learns both the structuring element and the composition of operators. It scales well to large datasets and online settings.	[ { "version": "v1", "created": "Tue, 11 Dec 2012 17:29:04 GMT" } ]	2012-12-12T00:00:00	[ [ "Masci", "Jonathan", "" ], [ "Angulo", "Jesús", "" ], [ "Schmidhuber", "Jürgen", "" ] ]	TITLE: A Learning Framework for Morphological Operators using Counter-Harmonic Mean ABSTRACT: We present a novel framework for learning morphological operators using counter-harmonic mean. It combines concepts from morphology and convolutional neural networks. A thorough experimental validation analyzes basic morphological operators dilation and erosion, opening and closing, as well as the much more complex top-hat transform, for which we report a real-world application from the steel industry. Using online learning and stochastic gradient descent, our system learns both the structuring element and the composition of operators. It scales well to large datasets and online settings.
1212.2573	K. S. Sesh Kumar	K. S. Sesh Kumar (LIENS, INRIA Paris - Rocquencourt), Francis Bach (LIENS, INRIA Paris - Rocquencourt)	Convex Relaxations for Learning Bounded Treewidth Decomposable Graphs	null	null	null	null	cs.LG cs.DS stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	We consider the problem of learning the structure of undirected graphical models with bounded treewidth, within the maximum likelihood framework. This is an NP-hard problem and most approaches consider local search techniques. In this paper, we pose it as a combinatorial optimization problem, which is then relaxed to a convex optimization problem that involves searching over the forest and hyperforest polytopes with special structures, independently. A supergradient method is used to solve the dual problem, with a run-time complexity of $O(k^3 n^{k+2} \log n)$ for each iteration, where $n$ is the number of variables and $k$ is a bound on the treewidth. We compare our approach to state-of-the-art methods on synthetic datasets and classical benchmarks, showing the gains of the novel convex approach.	[ { "version": "v1", "created": "Tue, 11 Dec 2012 18:22:31 GMT" } ]	2012-12-12T00:00:00	[ [ "Kumar", "K. S. Sesh", "", "LIENS, INRIA Paris - Rocquencourt" ], [ "Bach", "Francis", "", "LIENS, INRIA Paris - Rocquencourt" ] ]	TITLE: Convex Relaxations for Learning Bounded Treewidth Decomposable Graphs ABSTRACT: We consider the problem of learning the structure of undirected graphical models with bounded treewidth, within the maximum likelihood framework. This is an NP-hard problem and most approaches consider local search techniques. In this paper, we pose it as a combinatorial optimization problem, which is then relaxed to a convex optimization problem that involves searching over the forest and hyperforest polytopes with special structures, independently. A supergradient method is used to solve the dual problem, with a run-time complexity of $O(k^3 n^{k+2} \log n)$ for each iteration, where $n$ is the number of variables and $k$ is a bound on the treewidth. We compare our approach to state-of-the-art methods on synthetic datasets and classical benchmarks, showing the gains of the novel convex approach.
1212.1909	Luay Nakhleh	Yun Yu and Luay Nakhleh	Fast Algorithms for Reconciliation under Hybridization and Incomplete Lineage Sorting	null	null	null	null	q-bio.PE cs.DS	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Reconciling a gene tree with a species tree is an important task that reveals much about the evolution of genes, genomes, and species, as well as about the molecular function of genes. A wide array of computational tools have been devised for this task under certain evolutionary events such as hybridization, gene duplication/loss, or incomplete lineage sorting. Work on reconciling gene tree with species phylogenies under two or more of these events have also begun to emerge. Our group recently devised both parsimony and probabilistic frameworks for reconciling a gene tree with a phylogenetic network, thus allowing for the detection of hybridization in the presence of incomplete lineage sorting. While the frameworks were general and could handle any topology, they are computationally intensive, rendering their application to large datasets infeasible. In this paper, we present two novel approaches to address the computational challenges of the two frameworks that are based on the concept of ancestral configurations. Our approaches still compute exact solutions while improving the computational time by up to five orders of magnitude. These substantial gains in speed scale the applicability of these unified reconciliation frameworks to much larger data sets. We discuss how the topological features of the gene tree and phylogenetic network may affect the performance of the new algorithms. We have implemented the algorithms in our PhyloNet software package, which is publicly available in open source.	[ { "version": "v1", "created": "Sun, 9 Dec 2012 18:12:55 GMT" } ]	2012-12-11T00:00:00	[ [ "Yu", "Yun", "" ], [ "Nakhleh", "Luay", "" ] ]	TITLE: Fast Algorithms for Reconciliation under Hybridization and Incomplete Lineage Sorting ABSTRACT: Reconciling a gene tree with a species tree is an important task that reveals much about the evolution of genes, genomes, and species, as well as about the molecular function of genes. A wide array of computational tools have been devised for this task under certain evolutionary events such as hybridization, gene duplication/loss, or incomplete lineage sorting. Work on reconciling gene tree with species phylogenies under two or more of these events have also begun to emerge. Our group recently devised both parsimony and probabilistic frameworks for reconciling a gene tree with a phylogenetic network, thus allowing for the detection of hybridization in the presence of incomplete lineage sorting. While the frameworks were general and could handle any topology, they are computationally intensive, rendering their application to large datasets infeasible. In this paper, we present two novel approaches to address the computational challenges of the two frameworks that are based on the concept of ancestral configurations. Our approaches still compute exact solutions while improving the computational time by up to five orders of magnitude. These substantial gains in speed scale the applicability of these unified reconciliation frameworks to much larger data sets. We discuss how the topological features of the gene tree and phylogenetic network may affect the performance of the new algorithms. We have implemented the algorithms in our PhyloNet software package, which is publicly available in open source.
1212.1936	Nicolas Boulanger-Lewandowski	Nicolas Boulanger-Lewandowski, Yoshua Bengio and Pascal Vincent	High-dimensional sequence transduction	null	null	null	null	cs.LG	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	We investigate the problem of transforming an input sequence into a high-dimensional output sequence in order to transcribe polyphonic audio music into symbolic notation. We introduce a probabilistic model based on a recurrent neural network that is able to learn realistic output distributions given the input and we devise an efficient algorithm to search for the global mode of that distribution. The resulting method produces musically plausible transcriptions even under high levels of noise and drastically outperforms previous state-of-the-art approaches on five datasets of synthesized sounds and real recordings, approximately halving the test error rate.	[ { "version": "v1", "created": "Sun, 9 Dec 2012 23:28:02 GMT" } ]	2012-12-11T00:00:00	[ [ "Boulanger-Lewandowski", "Nicolas", "" ], [ "Bengio", "Yoshua", "" ], [ "Vincent", "Pascal", "" ] ]	TITLE: High-dimensional sequence transduction ABSTRACT: We investigate the problem of transforming an input sequence into a high-dimensional output sequence in order to transcribe polyphonic audio music into symbolic notation. We introduce a probabilistic model based on a recurrent neural network that is able to learn realistic output distributions given the input and we devise an efficient algorithm to search for the global mode of that distribution. The resulting method produces musically plausible transcriptions even under high levels of noise and drastically outperforms previous state-of-the-art approaches on five datasets of synthesized sounds and real recordings, approximately halving the test error rate.
1212.1633	Andrei Bulatov	Cong Wang and Andrei A. Bulatov	Inferring Attitude in Online Social Networks Based On Quadratic Correlation	18 pages, 3 figures	null	null	null	cs.SI physics.soc-ph	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	The structure of an online social network in most cases cannot be described just by links between its members. We study online social networks, in which members may have certain attitude, positive or negative toward each other, and so the network consists of a mixture of both positive and negative relationships. Our goal is to predict the sign of a given relationship based on the evidences provided in the current snapshot of the network. More precisely, using machine learning techniques we develop a model that after being trained on a particular network predicts the sign of an unknown or hidden link. The model uses relationships and influences from peers as evidences for the guess, however, the set of peers used is not predefined but rather learned during the training process. We use quadratic correlation between peer members to train the predictor. The model is tested on popular online datasets such as Epinions, Slashdot, and Wikipedia. In many cases it shows almost perfect prediction accuracy. Moreover, our model can also be efficiently updated as the underlaying social network evolves.	[ { "version": "v1", "created": "Fri, 7 Dec 2012 15:45:35 GMT" } ]	2012-12-10T00:00:00	[ [ "Wang", "Cong", "" ], [ "Bulatov", "Andrei A.", "" ] ]	TITLE: Inferring Attitude in Online Social Networks Based On Quadratic Correlation ABSTRACT: The structure of an online social network in most cases cannot be described just by links between its members. We study online social networks, in which members may have certain attitude, positive or negative toward each other, and so the network consists of a mixture of both positive and negative relationships. Our goal is to predict the sign of a given relationship based on the evidences provided in the current snapshot of the network. More precisely, using machine learning techniques we develop a model that after being trained on a particular network predicts the sign of an unknown or hidden link. The model uses relationships and influences from peers as evidences for the guess, however, the set of peers used is not predefined but rather learned during the training process. We use quadratic correlation between peer members to train the predictor. The model is tested on popular online datasets such as Epinions, Slashdot, and Wikipedia. In many cases it shows almost perfect prediction accuracy. Moreover, our model can also be efficiently updated as the underlaying social network evolves.
1211.6086	Kang Zhao	Kang Zhao, Greta Greer, Baojun Qiu, Prasenjit Mitra, Kenneth Portier, and John Yen	Finding influential users of an online health community: a new metric based on sentiment influence	Working paper	null	null	null	cs.SI cs.CY physics.soc-ph	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	What characterizes influential users in online health communities (OHCs)? We hypothesize that (1) the emotional support received by OHC members can be assessed from their sentiment ex-pressed in online interactions, and (2) such assessments can help to identify influential OHC members. Through text mining and sentiment analysis of users' online interactions, we propose a novel metric that directly measures a user's ability to affect the sentiment of others. Using dataset from an OHC, we demonstrate that this metric is highly effective in identifying influential users. In addition, combining the metric with other traditional measures further improves the identification of influential users. This study can facilitate online community management and advance our understanding of social influence in OHCs.	[ { "version": "v1", "created": "Mon, 26 Nov 2012 20:37:00 GMT" }, { "version": "v2", "created": "Wed, 5 Dec 2012 23:12:05 GMT" } ]	2012-12-07T00:00:00	[ [ "Zhao", "Kang", "" ], [ "Greer", "Greta", "" ], [ "Qiu", "Baojun", "" ], [ "Mitra", "Prasenjit", "" ], [ "Portier", "Kenneth", "" ], [ "Yen", "John", "" ] ]	TITLE: Finding influential users of an online health community: a new metric based on sentiment influence ABSTRACT: What characterizes influential users in online health communities (OHCs)? We hypothesize that (1) the emotional support received by OHC members can be assessed from their sentiment ex-pressed in online interactions, and (2) such assessments can help to identify influential OHC members. Through text mining and sentiment analysis of users' online interactions, we propose a novel metric that directly measures a user's ability to affect the sentiment of others. Using dataset from an OHC, we demonstrate that this metric is highly effective in identifying influential users. In addition, combining the metric with other traditional measures further improves the identification of influential users. This study can facilitate online community management and advance our understanding of social influence in OHCs.
1212.0888	Roozbeh Rajabi	Roozbeh Rajabi, Hassan Ghassemian	Unmixing of Hyperspectral Data Using Robust Statistics-based NMF	4 pages, conference	null	null	null	cs.CV	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Mixed pixels are presented in hyperspectral images due to low spatial resolution of hyperspectral sensors. Spectral unmixing decomposes mixed pixels spectra into endmembers spectra and abundance fractions. In this paper using of robust statistics-based nonnegative matrix factorization (RNMF) for spectral unmixing of hyperspectral data is investigated. RNMF uses a robust cost function and iterative updating procedure, so is not sensitive to outliers. This method has been applied to simulated data using USGS spectral library, AVIRIS and ROSIS datasets. Unmixing results are compared to traditional NMF method based on SAD and AAD measures. Results demonstrate that this method can be used efficiently for hyperspectral unmixing purposes.	[ { "version": "v1", "created": "Tue, 4 Dec 2012 21:59:35 GMT" } ]	2012-12-06T00:00:00	[ [ "Rajabi", "Roozbeh", "" ], [ "Ghassemian", "Hassan", "" ] ]	TITLE: Unmixing of Hyperspectral Data Using Robust Statistics-based NMF ABSTRACT: Mixed pixels are presented in hyperspectral images due to low spatial resolution of hyperspectral sensors. Spectral unmixing decomposes mixed pixels spectra into endmembers spectra and abundance fractions. In this paper using of robust statistics-based nonnegative matrix factorization (RNMF) for spectral unmixing of hyperspectral data is investigated. RNMF uses a robust cost function and iterative updating procedure, so is not sensitive to outliers. This method has been applied to simulated data using USGS spectral library, AVIRIS and ROSIS datasets. Unmixing results are compared to traditional NMF method based on SAD and AAD measures. Results demonstrate that this method can be used efficiently for hyperspectral unmixing purposes.
1212.1037	Tushar Rao Mr.	Tushar Rao (NSIT-Delhi) and Saket Srivastava (IIIT-Delhi)	Modeling Movements in Oil, Gold, Forex and Market Indices using Search Volume Index and Twitter Sentiments	10 pages, 4 figures, 9 Tables	null	null	IIITD-TR-2012-005	cs.CE cs.SI q-fin.GN	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Study of the forecasting models using large scale microblog discussions and the search behavior data can provide a good insight for better understanding the market movements. In this work we collected a dataset of 2 million tweets and search volume index (SVI from Google) for a period of June 2010 to September 2011. We perform a study over a set of comprehensive causative relationships and developed a unified approach to a model for various market securities like equity (Dow Jones Industrial Average-DJIA and NASDAQ-100), commodity markets (oil and gold) and Euro Forex rates. We also investigate the lagged and statistically causative relations of Twitter sentiments developed during active trading days and market inactive days in combination with the search behavior of public before any change in the prices/ indices. Our results show extent of lagged significance with high correlation value upto 0.82 between search volumes and gold price in USD. We find weekly accuracy in direction (up and down prediction) uptil 94.3% for DJIA and 90% for NASDAQ-100 with significant reduction in mean average percentage error for all the forecasting models.	[ { "version": "v1", "created": "Wed, 5 Dec 2012 14:28:40 GMT" } ]	2012-12-06T00:00:00	[ [ "Rao", "Tushar", "", "NSIT-Delhi" ], [ "Srivastava", "Saket", "", "IIIT-Delhi" ] ]	TITLE: Modeling Movements in Oil, Gold, Forex and Market Indices using Search Volume Index and Twitter Sentiments ABSTRACT: Study of the forecasting models using large scale microblog discussions and the search behavior data can provide a good insight for better understanding the market movements. In this work we collected a dataset of 2 million tweets and search volume index (SVI from Google) for a period of June 2010 to September 2011. We perform a study over a set of comprehensive causative relationships and developed a unified approach to a model for various market securities like equity (Dow Jones Industrial Average-DJIA and NASDAQ-100), commodity markets (oil and gold) and Euro Forex rates. We also investigate the lagged and statistically causative relations of Twitter sentiments developed during active trading days and market inactive days in combination with the search behavior of public before any change in the prices/ indices. Our results show extent of lagged significance with high correlation value upto 0.82 between search volumes and gold price in USD. We find weekly accuracy in direction (up and down prediction) uptil 94.3% for DJIA and 90% for NASDAQ-100 with significant reduction in mean average percentage error for all the forecasting models.
1212.1100	Jim Smith Dr	J. E. Smith, P. Caleb-Solly, M. A. Tahir, D. Sannen, H. van-Brussel	Making Early Predictions of the Accuracy of Machine Learning Applications	35 pagers, 12 figures	null	null	null	cs.LG cs.AI stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	The accuracy of machine learning systems is a widely studied research topic. Established techniques such as cross-validation predict the accuracy on unseen data of the classifier produced by applying a given learning method to a given training data set. However, they do not predict whether incurring the cost of obtaining more data and undergoing further training will lead to higher accuracy. In this paper we investigate techniques for making such early predictions. We note that when a machine learning algorithm is presented with a training set the classifier produced, and hence its error, will depend on the characteristics of the algorithm, on training set's size, and also on its specific composition. In particular we hypothesise that if a number of classifiers are produced, and their observed error is decomposed into bias and variance terms, then although these components may behave differently, their behaviour may be predictable. We test our hypothesis by building models that, given a measurement taken from the classifier created from a limited number of samples, predict the values that would be measured from the classifier produced when the full data set is presented. We create separate models for bias, variance and total error. Our models are built from the results of applying ten different machine learning algorithms to a range of data sets, and tested with "unseen" algorithms and datasets. We analyse the results for various numbers of initial training samples, and total dataset sizes. Results show that our predictions are very highly correlated with the values observed after undertaking the extra training. Finally we consider the more complex case where an ensemble of heterogeneous classifiers is trained, and show how we can accurately estimate an upper bound on the accuracy achievable after further training.	[ { "version": "v1", "created": "Wed, 5 Dec 2012 17:07:39 GMT" } ]	2012-12-06T00:00:00	[ [ "Smith", "J. E.", "" ], [ "Caleb-Solly", "P.", "" ], [ "Tahir", "M. A.", "" ], [ "Sannen", "D.", "" ], [ "van-Brussel", "H.", "" ] ]	TITLE: Making Early Predictions of the Accuracy of Machine Learning Applications ABSTRACT: The accuracy of machine learning systems is a widely studied research topic. Established techniques such as cross-validation predict the accuracy on unseen data of the classifier produced by applying a given learning method to a given training data set. However, they do not predict whether incurring the cost of obtaining more data and undergoing further training will lead to higher accuracy. In this paper we investigate techniques for making such early predictions. We note that when a machine learning algorithm is presented with a training set the classifier produced, and hence its error, will depend on the characteristics of the algorithm, on training set's size, and also on its specific composition. In particular we hypothesise that if a number of classifiers are produced, and their observed error is decomposed into bias and variance terms, then although these components may behave differently, their behaviour may be predictable. We test our hypothesis by building models that, given a measurement taken from the classifier created from a limited number of samples, predict the values that would be measured from the classifier produced when the full data set is presented. We create separate models for bias, variance and total error. Our models are built from the results of applying ten different machine learning algorithms to a range of data sets, and tested with "unseen" algorithms and datasets. We analyse the results for various numbers of initial training samples, and total dataset sizes. Results show that our predictions are very highly correlated with the values observed after undertaking the extra training. Finally we consider the more complex case where an ensemble of heterogeneous classifiers is trained, and show how we can accurately estimate an upper bound on the accuracy achievable after further training.
1212.1131	Lior Rokach	Gilad Katz, Guy Shani, Bracha Shapira, Lior Rokach	Using Wikipedia to Boost SVD Recommender Systems	null	null	null	null	cs.LG cs.IR stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Singular Value Decomposition (SVD) has been used successfully in recent years in the area of recommender systems. In this paper we present how this model can be extended to consider both user ratings and information from Wikipedia. By mapping items to Wikipedia pages and quantifying their similarity, we are able to use this information in order to improve recommendation accuracy, especially when the sparsity is high. Another advantage of the proposed approach is the fact that it can be easily integrated into any other SVD implementation, regardless of additional parameters that may have been added to it. Preliminary experimental results on the MovieLens dataset are encouraging.	[ { "version": "v1", "created": "Wed, 5 Dec 2012 19:03:39 GMT" } ]	2012-12-06T00:00:00	[ [ "Katz", "Gilad", "" ], [ "Shani", "Guy", "" ], [ "Shapira", "Bracha", "" ], [ "Rokach", "Lior", "" ] ]	TITLE: Using Wikipedia to Boost SVD Recommender Systems ABSTRACT: Singular Value Decomposition (SVD) has been used successfully in recent years in the area of recommender systems. In this paper we present how this model can be extended to consider both user ratings and information from Wikipedia. By mapping items to Wikipedia pages and quantifying their similarity, we are able to use this information in order to improve recommendation accuracy, especially when the sparsity is high. Another advantage of the proposed approach is the fact that it can be easily integrated into any other SVD implementation, regardless of additional parameters that may have been added to it. Preliminary experimental results on the MovieLens dataset are encouraging.
1212.0763	Modou Gueye M.	Modou Gueye, Talel Abdessalem, Hubert Naacke	Dynamic recommender system : using cluster-based biases to improve the accuracy of the predictions	31 pages, 7 figures	null	null	null	cs.LG cs.DB cs.IR	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	It is today accepted that matrix factorization models allow a high quality of rating prediction in recommender systems. However, a major drawback of matrix factorization is its static nature that results in a progressive declining of the accuracy of the predictions after each factorization. This is due to the fact that the new obtained ratings are not taken into account until a new factorization is computed, which can not be done very often because of the high cost of matrix factorization. In this paper, aiming at improving the accuracy of recommender systems, we propose a cluster-based matrix factorization technique that enables online integration of new ratings. Thus, we significantly enhance the obtained predictions between two matrix factorizations. We use finer-grained user biases by clustering similar items into groups, and allocating in these groups a bias to each user. The experiments we did on large datasets demonstrated the efficiency of our approach.	[ { "version": "v1", "created": "Mon, 3 Dec 2012 13:00:27 GMT" } ]	2012-12-05T00:00:00	[ [ "Gueye", "Modou", "" ], [ "Abdessalem", "Talel", "" ], [ "Naacke", "Hubert", "" ] ]	TITLE: Dynamic recommender system : using cluster-based biases to improve the accuracy of the predictions ABSTRACT: It is today accepted that matrix factorization models allow a high quality of rating prediction in recommender systems. However, a major drawback of matrix factorization is its static nature that results in a progressive declining of the accuracy of the predictions after each factorization. This is due to the fact that the new obtained ratings are not taken into account until a new factorization is computed, which can not be done very often because of the high cost of matrix factorization. In this paper, aiming at improving the accuracy of recommender systems, we propose a cluster-based matrix factorization technique that enables online integration of new ratings. Thus, we significantly enhance the obtained predictions between two matrix factorizations. We use finer-grained user biases by clustering similar items into groups, and allocating in these groups a bias to each user. The experiments we did on large datasets demonstrated the efficiency of our approach.
1212.0030	Andrew Habib	Osama Khalil, Andrew Habib	Viewpoint Invariant Object Detector	null	null	null	null	cs.CV	http://creativecommons.org/licenses/by/3.0/	Object Detection is the task of identifying the existence of an object class instance and locating it within an image. Difficulties in handling high intra-class variations constitute major obstacles to achieving high performance on standard benchmark datasets (scale, viewpoint, lighting conditions and orientation variations provide good examples). Suggested model aims at providing more robustness to detecting objects suffering severe distortion due to < 60{\deg} viewpoint changes. In addition, several model computational bottlenecks have been resolved leading to a significant increase in the model performance (speed and space) without compromising the resulting accuracy. Finally, we produced two illustrative applications showing the potential of the object detection technology being deployed in real life applications; namely content-based image search and content-based video search.	[ { "version": "v1", "created": "Fri, 30 Nov 2012 22:35:19 GMT" } ]	2012-12-04T00:00:00	[ [ "Khalil", "Osama", "" ], [ "Habib", "Andrew", "" ] ]	TITLE: Viewpoint Invariant Object Detector ABSTRACT: Object Detection is the task of identifying the existence of an object class instance and locating it within an image. Difficulties in handling high intra-class variations constitute major obstacles to achieving high performance on standard benchmark datasets (scale, viewpoint, lighting conditions and orientation variations provide good examples). Suggested model aims at providing more robustness to detecting objects suffering severe distortion due to < 60{\deg} viewpoint changes. In addition, several model computational bottlenecks have been resolved leading to a significant increase in the model performance (speed and space) without compromising the resulting accuracy. Finally, we produced two illustrative applications showing the potential of the object detection technology being deployed in real life applications; namely content-based image search and content-based video search.
1212.0087	Nader Jelassi	Mohamed Nader Jelassi and Sadok Ben Yahia and Engelbert Mephu Nguifo	A scalable mining of frequent quadratic concepts in d-folksonomies	null	null	null	null	cs.SI	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Folksonomy mining is grasping the interest of web 2.0 community since it represents the core data of social resource sharing systems. However, a scrutiny of the related works interested in mining folksonomies unveils that the time stamp dimension has not been considered. For example, the wealthy number of works dedicated to mining tri-concepts from folksonomies did not take into account time dimension. In this paper, we will consider a folksonomy commonly composed of triples <users, tags, resources> and we shall consider the time as a new dimension. We motivate our approach by highlighting the battery of potential applications. Then, we present the foundations for mining quadri-concepts, provide a formal definition of the problem and introduce a new efficient algorithm, called QUADRICONS for its solution to allow for mining folksonomies in time, i.e., d-folksonomies. We also introduce a new closure operator that splits the induced search space into equivalence classes whose smallest elements are the quadri-minimal generators. Carried out experiments on large-scale real-world datasets highlight good performances of our algorithm.	[ { "version": "v1", "created": "Sat, 1 Dec 2012 09:16:35 GMT" } ]	2012-12-04T00:00:00	[ [ "Jelassi", "Mohamed Nader", "" ], [ "Yahia", "Sadok Ben", "" ], [ "Nguifo", "Engelbert Mephu", "" ] ]	TITLE: A scalable mining of frequent quadratic concepts in d-folksonomies ABSTRACT: Folksonomy mining is grasping the interest of web 2.0 community since it represents the core data of social resource sharing systems. However, a scrutiny of the related works interested in mining folksonomies unveils that the time stamp dimension has not been considered. For example, the wealthy number of works dedicated to mining tri-concepts from folksonomies did not take into account time dimension. In this paper, we will consider a folksonomy commonly composed of triples <users, tags, resources> and we shall consider the time as a new dimension. We motivate our approach by highlighting the battery of potential applications. Then, we present the foundations for mining quadri-concepts, provide a formal definition of the problem and introduce a new efficient algorithm, called QUADRICONS for its solution to allow for mining folksonomies in time, i.e., d-folksonomies. We also introduce a new closure operator that splits the induced search space into equivalence classes whose smallest elements are the quadri-minimal generators. Carried out experiments on large-scale real-world datasets highlight good performances of our algorithm.
1212.0141	Yiye Ruan	Hemant Purohit and Yiye Ruan and David Fuhry and Srinivasan Parthasarathy and Amit Sheth	On the Role of Social Identity and Cohesion in Characterizing Online Social Communities	null	null	null	null	cs.SI physics.soc-ph	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Two prevailing theories for explaining social group or community structure are cohesion and identity. The social cohesion approach posits that social groups arise out of an aggregation of individuals that have mutual interpersonal attraction as they share common characteristics. These characteristics can range from common interests to kinship ties and from social values to ethnic backgrounds. In contrast, the social identity approach posits that an individual is likely to join a group based on an intrinsic self-evaluation at a cognitive or perceptual level. In other words group members typically share an awareness of a common category membership. In this work we seek to understand the role of these two contrasting theories in explaining the behavior and stability of social communities in Twitter. A specific focal point of our work is to understand the role of these theories in disparate contexts ranging from disaster response to socio-political activism. We extract social identity and social cohesion features-of-interest for large scale datasets of five real-world events and examine the effectiveness of such features in capturing behavioral characteristics and the stability of groups. We also propose a novel measure of social group sustainability based on the divergence in group discussion. Our main findings are: 1) Sharing of social identities (especially physical location) among group members has a positive impact on group sustainability, 2) Structural cohesion (represented by high group density and low average shortest path length) is a strong indicator of group sustainability, and 3) Event characteristics play a role in shaping group sustainability, as social groups in transient events behave differently from groups in events that last longer.	[ { "version": "v1", "created": "Sat, 1 Dec 2012 18:03:33 GMT" } ]	2012-12-04T00:00:00	[ [ "Purohit", "Hemant", "" ], [ "Ruan", "Yiye", "" ], [ "Fuhry", "David", "" ], [ "Parthasarathy", "Srinivasan", "" ], [ "Sheth", "Amit", "" ] ]	TITLE: On the Role of Social Identity and Cohesion in Characterizing Online Social Communities ABSTRACT: Two prevailing theories for explaining social group or community structure are cohesion and identity. The social cohesion approach posits that social groups arise out of an aggregation of individuals that have mutual interpersonal attraction as they share common characteristics. These characteristics can range from common interests to kinship ties and from social values to ethnic backgrounds. In contrast, the social identity approach posits that an individual is likely to join a group based on an intrinsic self-evaluation at a cognitive or perceptual level. In other words group members typically share an awareness of a common category membership. In this work we seek to understand the role of these two contrasting theories in explaining the behavior and stability of social communities in Twitter. A specific focal point of our work is to understand the role of these theories in disparate contexts ranging from disaster response to socio-political activism. We extract social identity and social cohesion features-of-interest for large scale datasets of five real-world events and examine the effectiveness of such features in capturing behavioral characteristics and the stability of groups. We also propose a novel measure of social group sustainability based on the divergence in group discussion. Our main findings are: 1) Sharing of social identities (especially physical location) among group members has a positive impact on group sustainability, 2) Structural cohesion (represented by high group density and low average shortest path length) is a strong indicator of group sustainability, and 3) Event characteristics play a role in shaping group sustainability, as social groups in transient events behave differently from groups in events that last longer.
1212.0146	Yiye Ruan	Yiye Ruan and David Fuhry and Srinivasan Parthasarathy	Efficient Community Detection in Large Networks using Content and Links	null	null	null	null	cs.SI physics.soc-ph	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	In this paper we discuss a very simple approach of combining content and link information in graph structures for the purpose of community discovery, a fundamental task in network analysis. Our approach hinges on the basic intuition that many networks contain noise in the link structure and that content information can help strengthen the community signal. This enables ones to eliminate the impact of noise (false positives and false negatives), which is particularly prevalent in online social networks and Web-scale information networks. Specifically we introduce a measure of signal strength between two nodes in the network by fusing their link strength with content similarity. Link strength is estimated based on whether the link is likely (with high probability) to reside within a community. Content similarity is estimated through cosine similarity or Jaccard coefficient. We discuss a simple mechanism for fusing content and link similarity. We then present a biased edge sampling procedure which retains edges that are locally relevant for each graph node. The resulting backbone graph can be clustered using standard community discovery algorithms such as Metis and Markov clustering. Through extensive experiments on multiple real-world datasets (Flickr, Wikipedia and CiteSeer) with varying sizes and characteristics, we demonstrate the effectiveness and efficiency of our methods over state-of-the-art learning and mining approaches several of which also attempt to combine link and content analysis for the purposes of community discovery. Specifically we always find a qualitative benefit when combining content with link analysis. Additionally our biased graph sampling approach realizes a quantitative benefit in that it is typically several orders of magnitude faster than competing approaches.	[ { "version": "v1", "created": "Sat, 1 Dec 2012 18:41:34 GMT" } ]	2012-12-04T00:00:00	[ [ "Ruan", "Yiye", "" ], [ "Fuhry", "David", "" ], [ "Parthasarathy", "Srinivasan", "" ] ]	TITLE: Efficient Community Detection in Large Networks using Content and Links ABSTRACT: In this paper we discuss a very simple approach of combining content and link information in graph structures for the purpose of community discovery, a fundamental task in network analysis. Our approach hinges on the basic intuition that many networks contain noise in the link structure and that content information can help strengthen the community signal. This enables ones to eliminate the impact of noise (false positives and false negatives), which is particularly prevalent in online social networks and Web-scale information networks. Specifically we introduce a measure of signal strength between two nodes in the network by fusing their link strength with content similarity. Link strength is estimated based on whether the link is likely (with high probability) to reside within a community. Content similarity is estimated through cosine similarity or Jaccard coefficient. We discuss a simple mechanism for fusing content and link similarity. We then present a biased edge sampling procedure which retains edges that are locally relevant for each graph node. The resulting backbone graph can be clustered using standard community discovery algorithms such as Metis and Markov clustering. Through extensive experiments on multiple real-world datasets (Flickr, Wikipedia and CiteSeer) with varying sizes and characteristics, we demonstrate the effectiveness and efficiency of our methods over state-of-the-art learning and mining approaches several of which also attempt to combine link and content analysis for the purposes of community discovery. Specifically we always find a qualitative benefit when combining content with link analysis. Additionally our biased graph sampling approach realizes a quantitative benefit in that it is typically several orders of magnitude faster than competing approaches.
1212.0317	M HM Krishna Prasad Dr	B. Adinarayana Reddy, O. Srinivasa Rao and M. H. M. Krishna Prasad	An Improved UP-Growth High Utility Itemset Mining	(0975 8887)	International Journal of Computer Applications Volume 58, No.2, 2012, 25-28	10.5120/9255-3424	null	cs.DB	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Efficient discovery of frequent itemsets in large datasets is a crucial task of data mining. In recent years, several approaches have been proposed for generating high utility patterns, they arise the problems of producing a large number of candidate itemsets for high utility itemsets and probably degrades mining performance in terms of speed and space. Recently proposed compact tree structure, viz., UP Tree, maintains the information of transactions and itemsets, facilitate the mining performance and avoid scanning original database repeatedly. In this paper, UP Tree (Utility Pattern Tree) is adopted, which scans database only twice to obtain candidate items and manage them in an efficient data structured way. Applying UP Tree to the UP Growth takes more execution time for Phase II. Hence this paper presents modified algorithm aiming to reduce the execution time by effectively identifying high utility itemsets.	[ { "version": "v1", "created": "Mon, 3 Dec 2012 08:50:50 GMT" } ]	2012-12-04T00:00:00	[ [ "Reddy", "B. Adinarayana", "" ], [ "Rao", "O. Srinivasa", "" ], [ "Prasad", "M. H. M. Krishna", "" ] ]	TITLE: An Improved UP-Growth High Utility Itemset Mining ABSTRACT: Efficient discovery of frequent itemsets in large datasets is a crucial task of data mining. In recent years, several approaches have been proposed for generating high utility patterns, they arise the problems of producing a large number of candidate itemsets for high utility itemsets and probably degrades mining performance in terms of speed and space. Recently proposed compact tree structure, viz., UP Tree, maintains the information of transactions and itemsets, facilitate the mining performance and avoid scanning original database repeatedly. In this paper, UP Tree (Utility Pattern Tree) is adopted, which scans database only twice to obtain candidate items and manage them in an efficient data structured way. Applying UP Tree to the UP Growth takes more execution time for Phase II. Hence this paper presents modified algorithm aiming to reduce the execution time by effectively identifying high utility itemsets.
1212.0402	Khurram Soomro	Khurram Soomro, Amir Roshan Zamir and Mubarak Shah	UCF101: A Dataset of 101 Human Actions Classes From Videos in The Wild	null	null	null	CRCV-TR-12-01	cs.CV	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	We introduce UCF101 which is currently the largest dataset of human actions. It consists of 101 action classes, over 13k clips and 27 hours of video data. The database consists of realistic user uploaded videos containing camera motion and cluttered background. Additionally, we provide baseline action recognition results on this new dataset using standard bag of words approach with overall performance of 44.5%. To the best of our knowledge, UCF101 is currently the most challenging dataset of actions due to its large number of classes, large number of clips and also unconstrained nature of such clips.	[ { "version": "v1", "created": "Mon, 3 Dec 2012 14:45:31 GMT" } ]	2012-12-04T00:00:00	[ [ "Soomro", "Khurram", "" ], [ "Zamir", "Amir Roshan", "" ], [ "Shah", "Mubarak", "" ] ]	TITLE: UCF101: A Dataset of 101 Human Actions Classes From Videos in The Wild ABSTRACT: We introduce UCF101 which is currently the largest dataset of human actions. It consists of 101 action classes, over 13k clips and 27 hours of video data. The database consists of realistic user uploaded videos containing camera motion and cluttered background. Additionally, we provide baseline action recognition results on this new dataset using standard bag of words approach with overall performance of 44.5%. To the best of our knowledge, UCF101 is currently the most challenging dataset of actions due to its large number of classes, large number of clips and also unconstrained nature of such clips.
1211.3375	Stephan Seufert	Stephan Seufert, Avishek Anand, Srikanta Bedathur, Gerhard Weikum	High-Performance Reachability Query Processing under Index Size Restrictions	30 pages	null	null	null	cs.DB cs.SI	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	In this paper, we propose a scalable and highly efficient index structure for the reachability problem over graphs. We build on the well-known node interval labeling scheme where the set of vertices reachable from a particular node is compactly encoded as a collection of node identifier ranges. We impose an explicit bound on the size of the index and flexibly assign approximate reachability ranges to nodes of the graph such that the number of index probes to answer a query is minimized. The resulting tunable index structure generates a better range labeling if the space budget is increased, thus providing a direct control over the trade off between index size and the query processing performance. By using a fast recursive querying method in conjunction with our index structure, we show that in practice, reachability queries can be answered in the order of microseconds on an off-the-shelf computer - even for the case of massive-scale real world graphs. Our claims are supported by an extensive set of experimental results using a multitude of benchmark and real-world web-scale graph datasets.	[ { "version": "v1", "created": "Wed, 14 Nov 2012 18:28:28 GMT" }, { "version": "v2", "created": "Mon, 19 Nov 2012 16:06:19 GMT" }, { "version": "v3", "created": "Mon, 26 Nov 2012 14:13:28 GMT" }, { "version": "v4", "created": "Wed, 28 Nov 2012 09:40:31 GMT" }, { "version": "v5", "created": "Thu, 29 Nov 2012 21:28:22 GMT" } ]	2012-12-03T00:00:00	[ [ "Seufert", "Stephan", "" ], [ "Anand", "Avishek", "" ], [ "Bedathur", "Srikanta", "" ], [ "Weikum", "Gerhard", "" ] ]	TITLE: High-Performance Reachability Query Processing under Index Size Restrictions ABSTRACT: In this paper, we propose a scalable and highly efficient index structure for the reachability problem over graphs. We build on the well-known node interval labeling scheme where the set of vertices reachable from a particular node is compactly encoded as a collection of node identifier ranges. We impose an explicit bound on the size of the index and flexibly assign approximate reachability ranges to nodes of the graph such that the number of index probes to answer a query is minimized. The resulting tunable index structure generates a better range labeling if the space budget is increased, thus providing a direct control over the trade off between index size and the query processing performance. By using a fast recursive querying method in conjunction with our index structure, we show that in practice, reachability queries can be answered in the order of microseconds on an off-the-shelf computer - even for the case of massive-scale real world graphs. Our claims are supported by an extensive set of experimental results using a multitude of benchmark and real-world web-scale graph datasets.
1211.6851	Chiheb-Eddine Ben n'cir C.B.N'cir	Chiheb-Eddine Ben N'Cir and Nadia Essoussi	Classification Recouvrante Bas\'ee sur les M\'ethodes \`a Noyau	Les 43\`emes Journ\'ees de Statistique	Les 43\`emes Journ\'ees de Statistique 2011	null	null	cs.LG stat.CO stat.ME stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Overlapping clustering problem is an important learning issue in which clusters are not mutually exclusive and each object may belongs simultaneously to several clusters. This paper presents a kernel based method that produces overlapping clusters on a high feature space using mercer kernel techniques to improve separability of input patterns. The proposed method, called OKM-K(Overlapping $k$-means based kernel method), extends OKM (Overlapping $k$-means) method to produce overlapping schemes. Experiments are performed on overlapping dataset and empirical results obtained with OKM-K outperform results obtained with OKM.	[ { "version": "v1", "created": "Thu, 29 Nov 2012 09:22:19 GMT" } ]	2012-11-30T00:00:00	[ [ "N'Cir", "Chiheb-Eddine Ben", "" ], [ "Essoussi", "Nadia", "" ] ]	TITLE: Classification Recouvrante Bas\'ee sur les M\'ethodes \`a Noyau ABSTRACT: Overlapping clustering problem is an important learning issue in which clusters are not mutually exclusive and each object may belongs simultaneously to several clusters. This paper presents a kernel based method that produces overlapping clusters on a high feature space using mercer kernel techniques to improve separability of input patterns. The proposed method, called OKM-K(Overlapping $k$-means based kernel method), extends OKM (Overlapping $k$-means) method to produce overlapping schemes. Experiments are performed on overlapping dataset and empirical results obtained with OKM-K outperform results obtained with OKM.
1211.6859	Chiheb-Eddine Ben n'cir C.B.N'cir	Chiheb-Eddine Ben N'Cir and Nadia Essoussi and Patrice Bertrand	Overlapping clustering based on kernel similarity metric	Second Meeting on Statistics and Data Mining 2010	Second Meeting on Statistics and Data Mining Second Meeting on Statistics and Data Mining March 11-12, 2010	null	null	stat.ML cs.LG stat.ME	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Producing overlapping schemes is a major issue in clustering. Recent proposed overlapping methods relies on the search of an optimal covering and are based on different metrics, such as Euclidean distance and I-Divergence, used to measure closeness between observations. In this paper, we propose the use of another measure for overlapping clustering based on a kernel similarity metric .We also estimate the number of overlapped clusters using the Gram matrix. Experiments on both Iris and EachMovie datasets show the correctness of the estimation of number of clusters and show that measure based on kernel similarity metric improves the precision, recall and f-measure in overlapping clustering.	[ { "version": "v1", "created": "Thu, 29 Nov 2012 09:35:30 GMT" } ]	2012-11-30T00:00:00	[ [ "N'Cir", "Chiheb-Eddine Ben", "" ], [ "Essoussi", "Nadia", "" ], [ "Bertrand", "Patrice", "" ] ]	TITLE: Overlapping clustering based on kernel similarity metric ABSTRACT: Producing overlapping schemes is a major issue in clustering. Recent proposed overlapping methods relies on the search of an optimal covering and are based on different metrics, such as Euclidean distance and I-Divergence, used to measure closeness between observations. In this paper, we propose the use of another measure for overlapping clustering based on a kernel similarity metric .We also estimate the number of overlapped clusters using the Gram matrix. Experiments on both Iris and EachMovie datasets show the correctness of the estimation of number of clusters and show that measure based on kernel similarity metric improves the precision, recall and f-measure in overlapping clustering.
1211.2881	Junyoung Chung	Junyoung Chung, Donghoon Lee, Youngjoo Seo, and Chang D. Yoo	Deep Attribute Networks	This paper has been withdrawn by the author due to a crucial grammatical errors	null	null	null	cs.CV cs.LG stat.ML	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Obtaining compact and discriminative features is one of the major challenges in many of the real-world image classification tasks such as face verification and object recognition. One possible approach is to represent input image on the basis of high-level features that carry semantic meaning which humans can understand. In this paper, a model coined deep attribute network (DAN) is proposed to address this issue. For an input image, the model outputs the attributes of the input image without performing any classification. The efficacy of the proposed model is evaluated on unconstrained face verification and real-world object recognition tasks using the LFW and the a-PASCAL datasets. We demonstrate the potential of deep learning for attribute-based classification by showing comparable results with existing state-of-the-art results. Once properly trained, the DAN is fast and does away with calculating low-level features which are maybe unreliable and computationally expensive.	[ { "version": "v1", "created": "Tue, 13 Nov 2012 03:41:31 GMT" }, { "version": "v2", "created": "Tue, 20 Nov 2012 11:30:46 GMT" }, { "version": "v3", "created": "Wed, 28 Nov 2012 08:39:03 GMT" } ]	2012-11-29T00:00:00	[ [ "Chung", "Junyoung", "" ], [ "Lee", "Donghoon", "" ], [ "Seo", "Youngjoo", "" ], [ "Yoo", "Chang D.", "" ] ]	TITLE: Deep Attribute Networks ABSTRACT: Obtaining compact and discriminative features is one of the major challenges in many of the real-world image classification tasks such as face verification and object recognition. One possible approach is to represent input image on the basis of high-level features that carry semantic meaning which humans can understand. In this paper, a model coined deep attribute network (DAN) is proposed to address this issue. For an input image, the model outputs the attributes of the input image without performing any classification. The efficacy of the proposed model is evaluated on unconstrained face verification and real-world object recognition tasks using the LFW and the a-PASCAL datasets. We demonstrate the potential of deep learning for attribute-based classification by showing comparable results with existing state-of-the-art results. Once properly trained, the DAN is fast and does away with calculating low-level features which are maybe unreliable and computationally expensive.
1208.3665	Christian Riess	Vincent Christlein, Christian Riess, Johannes Jordan, Corinna Riess and Elli Angelopoulou	An Evaluation of Popular Copy-Move Forgery Detection Approaches	Main paper: 14 pages, supplemental material: 12 pages, main paper appeared in IEEE Transaction on Information Forensics and Security	IEEE Transactions on Information Forensics and Security, volume 7, number 6, 2012, pp. 1841-1854	10.1109/TIFS.2012.2218597	null	cs.CV	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	A copy-move forgery is created by copying and pasting content within the same image, and potentially post-processing it. In recent years, the detection of copy-move forgeries has become one of the most actively researched topics in blind image forensics. A considerable number of different algorithms have been proposed focusing on different types of postprocessed copies. In this paper, we aim to answer which copy-move forgery detection algorithms and processing steps (e.g., matching, filtering, outlier detection, affine transformation estimation) perform best in various postprocessing scenarios. The focus of our analysis is to evaluate the performance of previously proposed feature sets. We achieve this by casting existing algorithms in a common pipeline. In this paper, we examined the 15 most prominent feature sets. We analyzed the detection performance on a per-image basis and on a per-pixel basis. We created a challenging real-world copy-move dataset, and a software framework for systematic image manipulation. Experiments show, that the keypoint-based features SIFT and SURF, as well as the block-based DCT, DWT, KPCA, PCA and Zernike features perform very well. These feature sets exhibit the best robustness against various noise sources and downsampling, while reliably identifying the copied regions.	[ { "version": "v1", "created": "Fri, 17 Aug 2012 19:41:23 GMT" }, { "version": "v2", "created": "Mon, 26 Nov 2012 20:53:51 GMT" } ]	2012-11-27T00:00:00	[ [ "Christlein", "Vincent", "" ], [ "Riess", "Christian", "" ], [ "Jordan", "Johannes", "" ], [ "Riess", "Corinna", "" ], [ "Angelopoulou", "Elli", "" ] ]	TITLE: An Evaluation of Popular Copy-Move Forgery Detection Approaches ABSTRACT: A copy-move forgery is created by copying and pasting content within the same image, and potentially post-processing it. In recent years, the detection of copy-move forgeries has become one of the most actively researched topics in blind image forensics. A considerable number of different algorithms have been proposed focusing on different types of postprocessed copies. In this paper, we aim to answer which copy-move forgery detection algorithms and processing steps (e.g., matching, filtering, outlier detection, affine transformation estimation) perform best in various postprocessing scenarios. The focus of our analysis is to evaluate the performance of previously proposed feature sets. We achieve this by casting existing algorithms in a common pipeline. In this paper, we examined the 15 most prominent feature sets. We analyzed the detection performance on a per-image basis and on a per-pixel basis. We created a challenging real-world copy-move dataset, and a software framework for systematic image manipulation. Experiments show, that the keypoint-based features SIFT and SURF, as well as the block-based DCT, DWT, KPCA, PCA and Zernike features perform very well. These feature sets exhibit the best robustness against various noise sources and downsampling, while reliably identifying the copied regions.
1211.5625	Sriganesh Srihari Dr	Sriganesh Srihari, Hon Wai Leong	A survey of computational methods for protein complex prediction from protein interaction networks	27 pages, 5 figures, 4 tables	Srihari, S., Leong, HW., J Bioinform Comput Biol 11(2): 1230002, 2013	10.1142/S021972001230002X	null	cs.CE q-bio.MN	http://arxiv.org/licenses/nonexclusive-distrib/1.0/	Complexes of physically interacting proteins are one of the fundamental functional units responsible for driving key biological mechanisms within the cell. Their identification is therefore necessary not only to understand complex formation but also the higher level organization of the cell. With the advent of high-throughput techniques in molecular biology, significant amount of physical interaction data has been cataloged from organisms such as yeast, which has in turn fueled computational approaches to systematically mine complexes from the network of physical interactions among proteins (PPI network). In this survey, we review, classify and evaluate some of the key computational methods developed till date for the identification of protein complexes from PPI networks. We present two insightful taxonomies that reflect how these methods have evolved over the years towards improving automated complex prediction. We also discuss some open challenges facing accurate reconstruction of complexes, the crucial ones being presence of high proportion of errors and noise in current high-throughput datasets and some key aspects overlooked by current complex detection methods. We hope this review will not only help to condense the history of computational complex detection for easy reference, but also provide valuable insights to drive further research in this area.	[ { "version": "v1", "created": "Sat, 24 Nov 2012 00:30:33 GMT" } ]	2012-11-27T00:00:00	[ [ "Srihari", "Sriganesh", "" ], [ "Leong", "Hon Wai", "" ] ]	TITLE: A survey of computational methods for protein complex prediction from protein interaction networks ABSTRACT: Complexes of physically interacting proteins are one of the fundamental functional units responsible for driving key biological mechanisms within the cell. Their identification is therefore necessary not only to understand complex formation but also the higher level organization of the cell. With the advent of high-throughput techniques in molecular biology, significant amount of physical interaction data has been cataloged from organisms such as yeast, which has in turn fueled computational approaches to systematically mine complexes from the network of physical interactions among proteins (PPI network). In this survey, we review, classify and evaluate some of the key computational methods developed till date for the identification of protein complexes from PPI networks. We present two insightful taxonomies that reflect how these methods have evolved over the years towards improving automated complex prediction. We also discuss some open challenges facing accurate reconstruction of complexes, the crucial ones being presence of high proportion of errors and noise in current high-throughput datasets and some key aspects overlooked by current complex detection methods. We hope this review will not only help to condense the history of computational complex detection for easy reference, but also provide valuable insights to drive further research in this area.

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.