Convergence of Uncertainty Sampling for Active Learning

TL;DR

This paper introduces a new uncertainty estimator for active learning in binary and multi-class classification, providing convergence guarantees under both noise-free and noisy conditions, advancing theoretical understanding of active learning algorithms.

Contribution

It proposes an efficient uncertainty estimator for binary and multi-class classification and offers the first non-asymptotic convergence guarantees for active learning with this estimator.

Findings

Provides a non-asymptotic convergence rate under no-noise conditions.

Extends analysis to noisy environments with theoretical guarantees.

Demonstrates effectiveness of the proposed estimator in active learning.

Abstract

Uncertainty sampling in active learning is heavily used in practice to reduce the annotation cost. However, there has been no wide consensus on the function to be used for uncertainty estimation in binary classification tasks and convergence guarantees of the corresponding active learning algorithms are not well understood. The situation is even more challenging for multi-category classification. In this work, we propose an efficient uncertainty estimator for binary classification which we also extend to multiple classes, and provide a non-asymptotic rate of convergence for our uncertainty sampling-based active learning algorithm in both cases under no-noise conditions (i.e., linearly separable data). We also extend our analysis to the noisy case and provide theoretical guarantees for our algorithm under the influence of noise in the task of binary and multi-class classification.