This repository contains a brief tutorial inspired by the paper "Wasserstein Auto-Encoders" by Tolstikhin, Bousquet, Gelly & Schölkopf (2017)
In this tutorial, we compare model frameworks for the generative adversarial network (GAN) formulation of the Wasserstein auto-encoder (WAEgan), the basic non-stochastic auto-encoder (AE), and the variational auto-encoder (VAE). To accomplish this, we implement each model in PyTorch as a convolutional auto-encoder similar to the popular DCGAN model and compare results with the MNIST and FashionMNIST datasets.
Contributors: