Hyperprior Induced Unsupervised Disentanglement of Latent Representations

TL;DR

This paper proposes a hierarchical Bayesian method using an inverse-Wishart prior to improve unsupervised disentanglement of latent representations in VAEs, outperforming existing methods on various datasets.

Contribution

It introduces a principled hierarchical Bayesian approach with an inverse-Wishart prior to control independence in latent spaces of VAEs, enhancing disentanglement.

Findings

Outperforms $eta$-VAE on multiple datasets

Achieves better disentanglement and reconstruction

Effective on datasets with correlated factors

Abstract

We address the problem of unsupervised disentanglement of latent representations learnt via deep generative models. In contrast to current approaches that operate on the evidence lower bound (ELBO), we argue that statistical independence in the latent space of VAEs can be enforced in a principled hierarchical Bayesian manner. To this effect, we augment the standard VAE with an inverse-Wishart (IW) prior on the covariance matrix of the latent code. By tuning the IW parameters, we are able to encourage (or discourage) independence in the learnt latent dimensions. Extensive experimental results on a range of datasets (2DShapes, 3DChairs, 3DFaces and CelebA) show our approach to outperform the $\beta$-VAE and is competitive with the state-of-the-art FactorVAE. Our approach achieves significantly better disentanglement and reconstruction on a new dataset (CorrelatedEllipses) which introduces correlations between the factors of variation.