An implementation of a variational autoencoder (VAE), see Kingma and Welling (2013). Useful resources I used to understand VAEs:
- Tutorial - What is a variational autoencoder? by Jaan Altosaar
- Variational Autoencoder: Intuition and Implementation by Augustinus Kristiadi
I embedded MNIST and Fashion MNIST in 2D space. The gifs are obtained by decoding points that lie on a circle in 2d space.