Formal Models of Active Learning from Contrastive Examples

TL;DR

This paper introduces a theoretical framework to analyze how contrastive examples influence active learning, focusing on sample complexity and revealing connections to self-directed learning, with applications to geometric and Boolean function classes.

Contribution

It provides a formal analysis of the impact of contrastive examples on active learning and links it to classical self-directed learning models.

Findings

Contrastive examples affect sample complexity in active learning.

A connection between contrastive learning and self-directed learning is established.

Results are demonstrated on geometric and Boolean function classes.

Abstract

Machine learning can greatly benefit from providing learning algorithms with pairs of contrastive training examples -- typically pairs of instances that differ only slightly, yet have different class labels. Intuitively, the difference in the instances helps explain the difference in the class labels. This paper proposes a theoretical framework in which the effect of various types of contrastive examples on active learners is studied formally. The focus is on the sample complexity of learning concept classes and how it is influenced by the choice of contrastive examples. We illustrate our results with geometric concept classes and classes of Boolean functions. Interestingly, we reveal a connection between learning from contrastive examples and the classical model of self-directed learning.