Enregistré dans:
Détails bibliographiques
Auteurs principaux: Gan, Feichen, Lu, Youcun, Zhang, Yingying, Liu, Yukun
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2510.26026
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866912678109249536
author Gan, Feichen
Lu, Youcun
Zhang, Yingying
Liu, Yukun
author_facet Gan, Feichen
Lu, Youcun
Zhang, Yingying
Liu, Yukun
contents Reliable uncertainty quantification is crucial for reinforcement learning (RL) in high-stakes settings. We propose a unified conformal prediction framework for infinite-horizon policy evaluation that constructs distribution-free prediction intervals {for returns} in both on-policy and off-policy settings. Our method integrates distributional RL with conformal calibration, addressing challenges such as unobserved returns, temporal dependencies, and distributional shifts. We propose a modular pseudo-return construction based on truncated rollouts and a time-aware calibration strategy using experience replay and weighted subsampling. These innovations mitigate model bias and restore approximate exchangeability, enabling uncertainty quantification even under policy shifts. Our theoretical analysis provides coverage guarantees that account for model misspecification and importance weight estimation. Empirical results, including experiments in synthetic and benchmark environments like Mountain Car, show that our method significantly improves coverage and reliability over standard distributional RL baselines.
format Preprint
id arxiv_https___arxiv_org_abs_2510_26026
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Conformal Prediction Beyond the Horizon: Distribution-Free Inference for Policy Evaluation
Gan, Feichen
Lu, Youcun
Zhang, Yingying
Liu, Yukun
Machine Learning
Reliable uncertainty quantification is crucial for reinforcement learning (RL) in high-stakes settings. We propose a unified conformal prediction framework for infinite-horizon policy evaluation that constructs distribution-free prediction intervals {for returns} in both on-policy and off-policy settings. Our method integrates distributional RL with conformal calibration, addressing challenges such as unobserved returns, temporal dependencies, and distributional shifts. We propose a modular pseudo-return construction based on truncated rollouts and a time-aware calibration strategy using experience replay and weighted subsampling. These innovations mitigate model bias and restore approximate exchangeability, enabling uncertainty quantification even under policy shifts. Our theoretical analysis provides coverage guarantees that account for model misspecification and importance weight estimation. Empirical results, including experiments in synthetic and benchmark environments like Mountain Car, show that our method significantly improves coverage and reliability over standard distributional RL baselines.
title Conformal Prediction Beyond the Horizon: Distribution-Free Inference for Policy Evaluation
topic Machine Learning
url https://arxiv.org/abs/2510.26026