t-SNE pytorch Implementation with CUDA

pytorch implementation of algorithm described in Visualizing Data using t-SNE. This code support cuda accelerating.

How to use it

Just download the repository, and the unzip mnist2500_X.zip or put feature file and labels file with code

1. run without cuda support

python tsne_torch.py --xfile mnist2500_X.txt --yfile mnist2500_labels.txt --cuda 0

2.run with cuda support

python tsne_torch.py --xfile mnist2500_X.txt --yfile mnist2500_labels.txt --cuda 1

Requirements

• pytorch
• matplotlib, numpy

Example

This is our result compare to result of python implementation.

• pytorch result

• python result

Credit

This code highly inspired by

• author's python implementation code here.