Multiview Representation Learning for a Union of Subspaces

TL;DR

This paper extends canonical correlation analysis (CCA) to a multiview mixture model, improving representation learning across multiple data views and demonstrating better downstream task performance.

Contribution

It introduces a novel multiview mixture model framework that generalizes CCA and offers simple heuristics for enhanced multiview representation learning.

Findings

Improved performance on downstream tasks using the proposed model.

The correlation-based objective generalizes standard CCA to mixtures.

Experimental results validate the effectiveness of the approach.

Abstract

Canonical correlation analysis (CCA) is a popular technique for learning representations that are maximally correlated across multiple views in data. In this paper, we extend the CCA based framework for learning a multiview mixture model. We show that the proposed model and a set of simple heuristics yield improvements over standard CCA, as measured in terms of performance on downstream tasks. Our experimental results show that our correlation-based objective meaningfully generalizes the CCA objective to a mixture of CCA models.