Evaluation of Active Feature Acquisition Methods for Time-varying Feature Settings
On the map: Offline RL and imitation learning from data → Offline RL algorithms
Abstract (arXiv)
Machine learning methods often assume that input features are available at no cost. However, in domains like healthcare, where acquiring features could be expensive or harmful, it is necessary to balance a feature's acquisition cost against its predictive value. The task of training an AI agent to decide which features to acquire is called active feature acquisition (AFA). By deploying an AFA agent, we effectively alter the acquisition strategy and trigger a distribution shift. To safely deploy AFA agents under this distribution shift, we present the problem of active feature acquisition performance evaluation (AFAPE). We examine AFAPE under i) a no direct effect (NDE) assumption, stating that acquisitions do not affect the underlying feature values; and ii) a no unobserved confounding (NUC) assumption, stating that retrospective feature acquisition decisions were only based on observed features. We show that one can apply missing data methods under the NDE assumption and offline reinforcement learning under the NUC assumption. When NUC and NDE hold, we propose a novel semi-offline reinforcement learning framework. This framework requires a weaker positivity assumption and introduces three new estimators: A direct method (DM), an inverse probability weighting (IPW), and a double reinforcement learning (DRL) estimator.
Key facts. Cited by 9 papers in the map (data through 7 Oct 2026). Placed in Offline RL algorithms, one of 13 directions in Offline RL and imitation learning from data. Builds on 22 of its 68 references that are in the map. First posted on arXiv 3 Dec 2023.
Builds on
22 of 68 references.
Closest ancestors · 1
Same direction, newest first
Other references · 21
- 2016Data-Efficient Off-Policy Policy Evaluation for Reinforcement Learningcited by 285
- 2022Semiparametric doubly robust targeted double machine learning: a reviewcited by 129
- 2019Double Reinforcement Learning for Efficient Off-Policy Evaluation in Markov Decision Processescited by 118
- 2018EDDI: Efficient Dynamic Discovery of High-Value Information with Partial VAEcited by 85
- 2021What's a good imputation to predict with missing values?cited by 61
- 2020Active Feature Acquisition with Generative Surrogate Modelscited by 53
- 2020Full Law Identification In Graphical Models Of Missing Data: Completeness Resultscited by 26
- 2019Classification with Costly Features as a Sequential Decision-Making Problemcited by 23
- 2020Reinforcement Learning with Efficient Active Feature Acquisitioncited by 20
- 2019Identification In Missing Data Models Represented By Directed Acyclic Graphscited by 19
- 2019ASAC: Active Sensing using Actor-Critic modelscited by 13
- 2020Inverse Active Sensing: Modeling and Understanding Timely Decision-Makingcited by 12
9 more
- 2018Active Feature Acquisition with Supervised Matrix Completioncited by 11
- 2017Semiparametric theorycited by 7
- 2018An Optimal Policy for Patient Laboratory Tests in Intensive Care Unitscited by 7
- 2019Dynamic Measurement Scheduling for Event Forecasting using Deep RLcited by 7
- 2020Dynamic Feature Acquisition with Arbitrary Conditional Flowscited by 6
- 2019Efficient estimation of optimal regimes under a no direct effect assumptioncited by 5
- 2021Towards Robust Active Feature Acquisitioncited by 4
- 2023From Single-Hospital to Multi-Centre Applications: Enhancing the Generalisability of Deep Learning Models for Adverse Event Prediction in the ICUcited by 2
- 2021Active Selection of Classification Featurescited by 1
Built on by
Citing papers per year: 2023: 1; 2024: 0; 2025: 4; 2026: 4.
9 papers.
- 2025A Survey on Active Feature Acquisition Strategiescited by 16
- 2023Evaluation of Active Feature Acquisition Methods for Static Feature Settingscited by 4
- 2025Cohort-Based Active Modality Acquisitioncited by 4
- 2025Learning-To-Measure: In-Context Active Feature Acquisitioncited by 4
- 2025Timely Clinical Diagnosis through Active Test Selectioncited by 4
- 2026Characterizing the Predictive Impact of Modalities with Supervised Latent-Variable Modelingcited by 0
- 2026BRiG-AFA: Bellman Risk-to-Go Learning for Non-Myopic Active Feature Acquisitioncited by 0
- 2026Counterfactual Evaluation of Temporal Observation Protocolscited by 0
- 2026When Does Test-Time Physical Diagnosis Pay? A Frozen Policy Buys Evidence It Never Readscited by 0
- 2026When Does Test-Time Physical Diagnosis Pay? A Frozen Policy Buys Evidence It Never Reads14 Sep 2026
- 2026Counterfactual Evaluation of Temporal Observation Protocols23 Aug 2026
- 2026BRiG-AFA: Bellman Risk-to-Go Learning for Non-Myopic Active Feature Acquisition3 Aug 2026
- 2026Characterizing the Predictive Impact of Modalities with Supervised Latent-Variable Modeling19 Feb 2026
- 2025Timely Clinical Diagnosis through Active Test Selection21 Oct 2025
- 2025Learning-To-Measure: In-Context Active Feature Acquisition14 Oct 2025
- 2025Cohort-Based Active Modality Acquisition22 May 2025
- 2025A Survey on Active Feature Acquisition Strategies16 Feb 2025
- 2023Evaluation of Active Feature Acquisition Methods for Static Feature Settings6 Dec 2023
Similar, not cited
Close in wording, no citation either way
- 2022Active Acquisition for Multimodal Temporal Data: A Challenging Decision-Making Task
- 2021Model Selection for Offline Reinforcement Learning: Practical Considerations for Healthcare Settings
- 2023Accountability in Offline Reinforcement Learning: Explaining Decisions with a Corpus of Examples
- 2021Measuring Data Quality for Dataset Selection in Offline Reinforcement Learning
- 2022DARA: Dynamics-Aware Reward Augmentation in Offline Reinforcement Learning
- 2025Beyond Non-Expert Demonstrations: Outcome-Driven Action Constraint for Offline Reinforcement Learning
- 2021Exploiting Action Impact Regularity and Exogenous State Variables for Offline Reinforcement Learning
- 2026Learning from the Unseen: Offline Reinforcement Learning with Hidden Actions