variational autoencoder

Variational Autoencoders (VAEs)[Kingma, et.al (2013)] let us design complex generative models of data that can be trained on large datasets. This post is about understanding the VAE concepts, its loss functions and how we can implement it in keras.