/VAE-Torch

Implementation of Variational Auto-Encoder in Torch7

Primary LanguageLuaMIT LicenseMIT

##Variational Auto-encoder

This is an improved implementation of the paper Stochastic Gradient VB and the Variational Auto-Encoder by D. Kingma and Prof. Dr. M. Welling. This code uses ReLUs and the adam optimizer, instead of sigmoids and adagrad. These changes make the network converge much faster.

In my other repository the implementation is in Python (Theano), this version is based on Torch7.

To run the MNIST experiment:

th main.lua

Setting the continuous boolean to true will make the script run the freyfaces experiment.

The code is MIT licensed.

I gratefully reused MNIST downloading and reading code written by Rahul G. Krishnan.