|
|
--- |
|
|
title: DictionaryLearningAndAnthropic |
|
|
emoji: 📈 |
|
|
colorFrom: blue |
|
|
colorTo: purple |
|
|
sdk: streamlit |
|
|
sdk_version: 1.35.0 |
|
|
app_file: app.py |
|
|
pinned: false |
|
|
license: mit |
|
|
--- |
|
|
|
|
|
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference |
|
|
|
|
|
Inspired by this video which discusses how Anthropic is studying how their model works using Dictionary Learning. |
|
|
https://youtu.be/AYTYk-kmXuI?t=168 |
|
|
|
|
|
|
|
|
Importing Libraries: |
|
|
|
|
|
streamlit for creating the app. |
|
|
Scikit-learn (sklearn.decomposition.DictionaryLearning and datasets.load_digits) for dictionary learning and loading the dataset. |
|
|
matplotlib for plotting. |
|
|
numpy for numerical operations. |
|
|
Streamlit App Initialization: |
|
|
|
|
|
The app sets a title and a description with st.title and st.write. |
|
|
Loading Data: |
|
|
|
|
|
Digit data is loaded using load_digits(). This dataset contains 8x8 images of digits (0-9). |
|
|
Sample Image Display: |
|
|
|
|
|
A slider is created with st.slider to select an index for displaying a sample image from the dataset using st.image. |
|
|
Selecting Number of Components: |
|
|
|
|
|
Another slider allows users to choose the number of dictionary components (n_components) for dictionary learning. |
|
|
Performing Dictionary Learning: |
|
|
|
|
|
DictionaryLearning is initialized with the selected number of components and fitted to the data. |
|
|
The dictionary elements learned are stored in dictionary. |
|
|
Displaying Dictionary Components: |
|
|
|
|
|
The learned dictionary components are displayed as images using Matplotlib. Each component is reshaped into an 8x8 image. |
|
|
Displaying Sparsity: |
|
|
|
|
|
Sparsity is calculated as the mean of zero entries in the transformed data (X_transformed) and displayed with st.write. |
|
|
|
|
|
|
|
|
Assistant: Sure! The requirements.txt file is used to specify the dependencies needed for your Python project. For the Streamlit application demonstrating dictionary learning, you'll need the following libraries: |
|
|
|
|
|
streamlit to create the interactive web app. |
|
|
scikit-learn for performing dictionary learning and loading the dataset. |
|
|
matplotlib for plotting the images. |
|
|
numpy for numerical operations. |
|
|
Here is a requirements.txt file that includes all these dependencies: |
|
|
|
|
|
streamlit>=1.0.0 |
|
|
scikit-learn>=0.24.0 |
|
|
matplotlib>=3.0.0 |
|
|
numpy>=1.19.0 |
|
|
|
|
|
|
|
|
|
|
|
|
