Information Constraints On Auto-Encoding Variational Bayes

Authors:
Romain Lopez UC Berkeley
Jeff Regier UC Berkeley
Michael I. Jordan UC Berkeley
Nir Yosef UC Berkeley

Introduction:

Parameterizing the approximate posterior of a generative model with neural networks has become a common theme in recent machine learning research.The authors 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.

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 in 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.

You may want to know: