Information Constraints on Auto-Encoding Variational Bayes

TL;DR

This paper introduces a framework that uses kernel-based independence measures within Auto-Encoding Variational Bayes to learn more interpretable and invariant representations, demonstrated on complex biological data.

Contribution

It proposes a novel method combining dHSIC with variational autoencoders to enforce independence constraints, improving interpretability and invariance in learned representations.

Findings

Outperforms state-of-the-art in single-cell RNA sequencing analysis

Enforces independence between latent variables and nuisance factors

Applicable to learning invariant and interpretable representations

Abstract

Parameterizing the approximate posterior of a generative model with neural networks has become a common theme in recent machine learning research. While providing appealing flexibility, this approach makes it difficult to impose or assess structural constraints such as conditional independence. We propose a framework for learning representations that relies on Auto-Encoding Variational Bayes and whose search space is constrained via kernel-based measures of independence. In particular, our method employs the $d$-variable Hilbert-Schmidt Independence Criterion (dHSIC) to enforce independence between the latent representations and arbitrary nuisance factors. We show how to apply this method to a range of problems, including the problems of learning invariant representations and the learning of interpretable representations. We also present a full-fledged application to single-cell RNA sequencing (scRNA-seq). In this setting the biological signal is mixed in complex ways with sequencing errors and sampling effects. We show that our method out-performs the state-of-the-art in this domain.