This project aims at unsupervised clustering through generative models. Thus a variational autoencoder is trained to cluster data in its encoder. Instead of an isotropic gaussian prior, the input is considered to be composed of a mixture of K gaussians, K being the number of clusters the data may posses.
This project is based on the paper:
Nat Dilokthanakul, Pedro A.M. Mediano, Marta Garnelo, Matthew C.H. Lee, Hugh Salimbeni, Kai Arulkumaran, Murray Shanahan. Deep Unsupervised Clustering with Gaussian Mixture Variational Autoencoders.
Link: https://arxiv.org/abs/1611.02648
This network on MNIST dataset achieves approximately 75% classification accuracy on test set after learning the clusters in training data.