Budgeted Online Active Learning with Expert Advice and Episodic Priors

2025-06-03

AI Summary

This work addresses online active learning under stringent constraints of limited time windows and extremely low annotation budgets—particularly relevant to high-cost domains such as agriculture. Existing approaches fail to jointly model annotation budget limitations, temporal constraints, and the periodic prior inherent in expert labeling behavior. To bridge this gap, we propose the first online active learning framework that unifies budget-aware querying, time-domain constraints, and periodic expert behavior modeling grounded in unlabeled data streams, integrating expert advice aggregation with streaming decision-making mechanisms. Evaluated on a crop simulator and real-world multi-cultivar grape datasets, our method significantly outperforms baseline experts, uniform sampling, and existing methods that ignore periodicity—even under ultra-sparse annotation budgets—while maintaining strong generalization. Results empirically validate that incorporating periodic priors is critical for efficient, high-utility query selection in resource-constrained online learning settings.

Technology Category

• Machine Learning: Active Learning
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs

Abstract

This paper introduces a novel approach to budgeted online active learning from finite-horizon data streams with extremely limited labeling budgets. In agricultural applications, such streams might include daily weather data over a growing season, and labels require costly measurements of weather-dependent plant characteristics. Our method integrates two key sources of prior information: a collection of preexisting expert predictors and episodic behavioral knowledge of the experts based on unlabeled data streams. Unlike previous research on online active learning with experts, our work simultaneously considers query budgets, finite horizons, and episodic knowledge, enabling effective learning in applications with severely limited labeling capacity. We demonstrate the utility of our approach through experiments on various prediction problems derived from both a realistic agricultural crop simulator and real-world data from multiple grape cultivars. The results show that our method significantly outperforms baseline expert predictions, uniform query selection, and existing approaches that consider budgets and limited horizons but neglect episodic knowledge, even under highly constrained labeling budgets.

Problem

• Budgeted online active learning with limited labeling budgets
• Integrating expert predictors and episodic behavioral knowledge
• Effective learning under severe labeling capacity constraints

Innovation

• Integrates expert predictors and episodic priors
• Considers query budgets and finite horizons
• Outperforms baselines with constrained labeling budgets