Multiview Learning of Weighted Majority Vote by Bregman Divergence Minimization

TL;DR

This paper introduces a multiview learning method that combines classifiers using Bregman divergence minimization, enabling effective classifier ensemble learning even with limited labeled data.

Contribution

It proposes a novel multiview weighted majority vote framework with a Bregman divergence-based optimization algorithm for classifier combination.

Findings

The method effectively leverages multiview data for classifier ensemble.

It overcomes limited labeled data issues in multiview learning.

The algorithm demonstrates competitive performance in experiments.

Abstract

We tackle the issue of classifier combinations when observations have multiple views. Our method jointly learns view-specific weighted majority vote classifiers (i.e. for each view) over a set of base voters, and a second weighted majority vote classifier over the set of these view-specific weighted majority vote classifiers. We show that the empirical risk minimization of the final majority vote given a multiview training set can be cast as the minimization of Bregman divergences. This allows us to derive a parallel-update optimization algorithm for learning our multiview model. We empirically study our algorithm with a particular focus on the impact of the training set size on the multiview learning results. The experiments show that our approach is able to overcome the lack of labeled information.