https://github.com/msminhas93/embeddings-visualization

https://github.com/msminhas93/embeddings-visualization

Last synced: 3 months ago
JSON representation

• Host: GitHub
• URL: https://github.com/msminhas93/embeddings-visualization
• Owner: msminhas93
• Created: 2021-01-31T03:14:14.000Z (almost 4 years ago)
• Default Branch: main
• Last Pushed: 2021-01-31T03:25:54.000Z (almost 4 years ago)
• Last Synced: 2024-11-01T06:42:05.404Z (3 months ago)
• Language: Jupyter Notebook
• Size: 1.93 MB
• Stars: 36
• Watchers: 1
• Forks: 10
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Embeddings / Feature Vector visualization using t-SNE and PCA

This repository contains the code for visualizing embeddings / feature vectors using t-SNE and PCA.The example plots are shown for model trained on the MNIST dataset with 1000 noise images randomly injected into the test set. The training code is also available. I used PyTorch and Lightning in this for the DL component. However, the logic should be similar for any framework.

The sample plots of embeddings extracted from a trained model are shown below.

t-SNE plot 2D

![t-SNE plot 2D](plots/TSNEtrained.png)

t-SNE plot 3D

![t-SNE plot 3D](plots/TSNE3D.png)

PCA plot 2D

![PCA plot 2D](plots/PCAtrained.png)

PCA plot 3D

![PCA plot 3D](plots/PCA3D.png)

I've shown the projections of embeddings extracted from randomly initialized model as well.

t-SNE plot 2D

![t-SNE plot 2D](plots/TSNErandom.png)

PCA plot 2D

![PCA plot 2D](plots/PCArandom.png)