Unsupervised Learning of Global Factors in Deep Generative Models

TL;DR

This paper introduces a new deep generative model that learns global dependencies in data without supervision, capturing interpretable representations, enabling domain alignment, and discriminating complex groups of observations.

Contribution

The proposed model uniquely combines a mixture model with a global Gaussian latent variable to learn global factors in an unsupervised manner, without regularization.

Findings

Latent global space captures disentangled, interpretable representations.

Model performs effective domain alignment and interpolation.

Global space discriminates complex structured groups of observations.

Abstract

We present a novel deep generative model based on non i.i.d. variational autoencoders that captures global dependencies among observations in a fully unsupervised fashion. In contrast to the recent semi-supervised alternatives for global modeling in deep generative models, our approach combines a mixture model in the local or data-dependent space and a global Gaussian latent variable, which lead us to obtain three particular insights. First, the induced latent global space captures interpretable disentangled representations with no user-defined regularization in the evidence lower bound (as in $\beta$-VAE and its generalizations). Second, we show that the model performs domain alignment to find correlations and interpolate between different databases. Finally, we study the ability of the global space to discriminate between groups of observations with non-trivial underlying structures, such as face images with shared attributes or defined sequences of digits images.