Information Templates: A New Paradigm for Intelligent Active Feature Acquisition

2025-08-25

AI Summary

Existing active feature acquisition (AFA) methods suffer from either the high computational complexity of reinforcement learning or the suboptimality of greedy strategies, which overly rely on joint feature informativeness and explicit data distribution modeling. To address these limitations, we propose Template-based Active Feature Acquisition (TAFA), the first AFA framework introducing a learnable feature template mechanism. TAFA models cross-instance joint informativeness to transform sequential feature selection into a template-guided sequential decision process—thereby obviating explicit data distribution estimation and RL policy search. As a result, TAFA achieves non-greedy selection, strong generalization across domains, and computational efficiency. Extensive experiments on multiple synthetic and real-world datasets demonstrate that TAFA significantly reduces both feature acquisition cost and computational overhead, consistently outperforming state-of-the-art methods.

Technology Category

• Machine Learning: Active Learning
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Active feature acquisition (AFA) is an instance-adaptive paradigm in which, at test time, a policy sequentially chooses which features to acquire (at a cost) before predicting. Existing approaches either train reinforcement learning (RL) policies, which deal with a difficult MDP, or greedy policies that cannot account for the joint informativeness of features or require knowledge about the underlying data distribution. To overcome this, we propose Template-based AFA (TAFA), a non-greedy framework that learns a small library of feature templates--a set of features that are jointly informative--and uses this library of templates to guide the next feature acquisitions. Through identifying feature templates, the proposed framework not only significantly reduces the action space considered by the policy but also alleviates the need to estimate the underlying data distribution. Extensive experiments on synthetic and real-world datasets show that TAFA outperforms the existing state-of-the-art baselines while achieving lower overall acquisition cost and computation.

Problem

• Reducing action space in active feature acquisition
• Eliminating need for data distribution estimation
• Lowering feature acquisition costs and computation

Innovation

• Template-based AFA framework
• Learns library of feature templates
• Reduces action space and distribution estimation