Adaptive Active Learning for Regression via Reinforcement Learning

TL;DR

This paper introduces WiGS, a reinforcement learning-based adaptive active learning method for regression that dynamically balances exploration and exploitation, outperforming existing static methods across diverse datasets.

Contribution

It proposes a novel reinforcement learning framework to optimize sample selection in active regression, replacing static rules with a dynamic, additive criterion.

Findings

WiGS outperforms iGS and baselines in accuracy and efficiency.

WiGS effectively handles irregular data densities.

Reinforcement learning enables adaptive exploration-exploitation balance.

Abstract

Active learning for regression reduces labeling costs by selecting the most informative samples. Improved Greedy Sampling is a prominent method that balances feature-space diversity and output-space uncertainty using a static, multiplicative rule. We propose Weighted improved Greedy Sampling (WiGS), which replaces this framework with a dynamic, additive criterion. We formulate weight selection as a reinforcement learning problem, enabling an agent to adapt the exploration-investigation balance throughout learning. Experiments on 18 benchmark datasets and a synthetic environment show WiGS outperforms iGS and other baseline methods in both accuracy and labeling efficiency, particularly in domains with irregular data density where the baseline's multiplicative rule ignores high-error samples in dense regions.