These are notebook are explaining the variational autoencoder (vae) described in Kingma&Welling [1312.6114]
Further details are given in a talk: https://home.zhaw.ch/~dueo/bbs/files/vae.pdf
- vae_demo-2D.ipynb is a toy example reproducing a two dimensional distribution from a one dimensional latent space
- vae_demo.ipynb reproduces the MNIST-Digits from a two dimensional latent space