Agnostic Active Learning Without Constraints

TL;DR

This paper introduces an agnostic active learning algorithm that operates without maintaining a version space, reducing computational complexity while still significantly improving classification performance over supervised learning.

Contribution

The paper proposes a novel active learning method that eliminates the need for version spaces, simplifying implementation and potentially enhancing robustness.

Findings

Achieves substantial improvements over supervised learning.

Operates without maintaining a version space, reducing computational burden.

Provides theoretical analysis of the algorithm's performance.

Abstract

We present and analyze an agnostic active learning algorithm that works without keeping a version space. This is unlike all previous approaches where a restricted set of candidate hypotheses is maintained throughout learning, and only hypotheses from this set are ever returned. By avoiding this version space approach, our algorithm sheds the computational burden and brittleness associated with maintaining version spaces, yet still allows for substantial improvements over supervised learning for classification.