VAE-SNE is a deep generative model for both dimensionality reduction and clustering. VAE-SNE is a variational autoencoder (VAE) regularized with the stochastic neighbor embedding (t-SNE/SNE) objective to improve local structure preservation in the compressed latent space. The model simultaneously learns a Gaussian mixture cluster distribution during optimization, and overlapping mixture components are then combined using a sparse watershed procedure, so the number of clusters does not have to be specified manually — provided the number of Gaussian mixture components is large enough. VAE-SNE produces embeddings with similar quality to existing dimensionality reduction methods; can detect outliers; scales to large, out-of-core datasets; and can easily add new data to an existing embedding/clustering.
Don't forget to tag @jgraving in your comment, otherwise they may not be notified.