Theoretical Foundation of Co-Training and Disagreement-Based Algorithms

TL;DR

This paper provides a theoretical foundation for disagreement-based semi-supervised learning methods, especially co-training, analyzing their principles and addressing unresolved issues to enhance understanding and application.

Contribution

It offers the first comprehensive theoretical analysis of co-training and disagreement-based algorithms, clarifying their underlying principles and resolving key open problems.

Findings

Theoretical validation of co-training effectiveness

Analysis of conditions for successful disagreement-based learning

Insights into the role of multiple classifiers and views

Abstract

Disagreement-based approaches generate multiple classifiers and exploit the disagreement among them with unlabeled data to improve learning performance. Co-training is a representative paradigm of them, which trains two classifiers separately on two sufficient and redundant views; while for the applications where there is only one view, several successful variants of co-training with two different classifiers on single-view data instead of two views have been proposed. For these disagreement-based approaches, there are several important issues which still are unsolved, in this article we present theoretical analyses to address these issues, which provides a theoretical foundation of co-training and disagreement-based approaches.