Teaching a Variational Autoencoder (VAE) to draw MNIST characters

These characters have not been written by a human — we taught a neural network how to do this!

To see the full VAE code, please refer to my github.

Autoencoders are a type of neural network that can be used to learn efficient codings of input data. Given some inputs, the network first applies a series of transformations that map the input data into a lower dimensional space. This part of the network is called the encoder.