Minimal Variational Auto-Encoder
This is a minimal implementation of an Variational Auto-Encoder in Tensorflow applied to MNIST.
Some example generated numbers:
How to run
Simply clone the directory and run the file
vae_mnist.py. Results will be displayed in real time, while full training takes a few minutes.
The implementation follows Auto-Encoding Variational Bayes. Both the generator and discriminator uses 3 convolutional layers with 5x5 convolutions, with obvious room for improvements.