CUDA Implementation of T-SNE with Python bindings
tsnecuda provides an optimized CUDA implementation of the T-SNE algorithm by L Van der Maaten. tsnecuda is able to compute the T-SNE of large numbers of points up to 1200 times faster than other leading libraries, and provides simple python bindings with a SKLearn style interface::
#!/usr/bin/env python from tsnecuda import TSNE embeddedX = TSNE(n_components=2).fit_transform(X)
For more information, check out the repository at https://github.com/rmrao/tsne-cuda.