This is an implementation of a variational auto-encoder for the MNIST dataset using PyTorch. The method was introduced in this paper by Kingma, Diederik P. and Max Welling.
The following figure shows the training loss and testing loss after each epoch of training.
The following figure shows the samples generated from the model.
This figure shows the generated samples in the 2D latent space.
This figure shows the embedding of training samples in 2D latent space. The points are colored by classes
