Skip to content

Latest commit

 

History

History
13 lines (9 loc) · 587 Bytes

README.mdown

File metadata and controls

13 lines (9 loc) · 587 Bytes

Another fairly minimal implementation of variational autoencoder (Kingma et al., ICLR 2014) in Theano.

Requirements: Theano, SciPy, and matplotlib.

Example training and visualization for MNIST dataset can be found in mnist_manifold.sh. You can grab the dataset here; then change the path in utils.py accordingly.

Example manifold generated:

AEVB implementation