VAE with a VampPrior

TL;DR

This paper introduces VampPrior, a novel mixture-based prior for VAEs, which improves model performance and avoids local optima, demonstrated through extensive empirical evaluation on multiple datasets.

Contribution

The paper proposes VampPrior, a new learnable mixture prior for VAEs, and extends it to a hierarchical model, achieving state-of-the-art results across various datasets.

Findings

VampPrior improves VAE performance on multiple datasets.

Hierarchical VampPrior learns better representations and avoids local optima.

Achieves state-of-the-art results in unsupervised learning tasks.

Abstract

Many different methods to train deep generative models have been introduced in the past. In this paper, we propose to extend the variational auto-encoder (VAE) framework with a new type of prior which we call "Variational Mixture of Posteriors" prior, or VampPrior for short. The VampPrior consists of a mixture distribution (e.g., a mixture of Gaussians) with components given by variational posteriors conditioned on learnable pseudo-inputs. We further extend this prior to a two layer hierarchical model and show that this architecture with a coupled prior and posterior, learns significantly better models. The model also avoids the usual local optima issues related to useless latent dimensions that plague VAEs. We provide empirical studies on six datasets, namely, static and binary MNIST, OMNIGLOT, Caltech 101 Silhouettes, Frey Faces and Histopathology patches, and show that applying the hierarchical VampPrior delivers state-of-the-art results on all datasets in the unsupervised permutation invariant setting and the best results or comparable to SOTA methods for the approach with convolutional networks.