Adaptive Active Learning for Regression via Reinforcement Learning

2026-03-11

AI Summary

This work addresses the inefficiency in sample selection for regression tasks caused by high annotation costs and non-uniform data density. The authors propose WiGS, a novel approach that formulates active learning–based sample selection as a reinforcement learning problem. Its key innovation lies in replacing conventional static multiplicative rules with a dynamic additive criterion, which adaptively balances exploration in the feature space against uncertainty in the output space. This is further enhanced by an improved greedy sampling strategy coupled with deep reinforcement learning to enable online optimization of weighting parameters. Extensive experiments across 18 benchmark and synthetic datasets demonstrate that WiGS consistently outperforms existing methods, achieving notable gains in both predictive accuracy and labeling efficiency—particularly under irregular data distributions.

Technology Category

• Machine Learning: Active Learning
• Search and Optimization: Sampling/Simulation-based Search
• Reasoning under Uncertainty: Stochastic Optimization

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

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.

Problem

• active learning
• regression
• sample selection
• labeling efficiency
• data density

Innovation

• Active Learning
• Regression
• Reinforcement Learning
• Sample Selection
• Dynamic Weighting