Categorical Variational Autoencoder

Keras, Tensorflow Probability and Eager Execution Implementation

Straight Through Gumbel-Softmax Estimator implemented as per paper: Categorical Reparameterization with Gumbel-Softmax (No temperature, learning rate annealing. Hard prior used)

Code developed from:

1. Eric Jang: https://github.com/ericjang/gumbel-softmax/blob/master/gumbel_softmax_vae_v2.ipynb

2. Google Seedbank Convolutional Variational Autoencoder https://tools.google.com/seedbank/seed/5719238044024832

File: cat_vae_v0.1.py - Sep 24, 2018

• Tensorflow 1.10.0
• Numpy 1.14.5
• Epochs = 10
• Temperature = 1
• Learning Rate = 0.001
• Number of categorical distributions = 30

Items for further development:

• Anneal temperature and learning rate
• Use relaxed prior
• Increase the number of epochs
• Increase the number of categorical distributions to sample from

Example of generated MNIST images from 100 test samples