Adaptive $Q$-Aid for Conditional Supervised Learning in Offline Reinforcement Learning

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Kim, Jeonghye, Lee, Suyoung, Kim, Woojun, Sung, Youngchul
Format: Preprint
Publié: 2024
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866915859876806656
author Kim, Jeonghye
Lee, Suyoung
Kim, Woojun
Sung, Youngchul
author_facet Kim, Jeonghye
Lee, Suyoung
Kim, Woojun
Sung, Youngchul
contents Offline reinforcement learning (RL) has progressed with return-conditioned supervised learning (RCSL), but its lack of stitching ability remains a limitation. We introduce $Q$-Aided Conditional Supervised Learning (QCS), which effectively combines the stability of RCSL with the stitching capability of $Q$-functions. By analyzing $Q$-function over-generalization, which impairs stable stitching, QCS adaptively integrates $Q$-aid into RCSL's loss function based on trajectory return. Empirical results show that QCS significantly outperforms RCSL and value-based methods, consistently achieving or exceeding the maximum trajectory returns across diverse offline RL benchmarks.
format Preprint
id arxiv_https___arxiv_org_abs_2402_02017
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Adaptive $Q$-Aid for Conditional Supervised Learning in Offline Reinforcement Learning
Kim, Jeonghye
Lee, Suyoung
Kim, Woojun
Sung, Youngchul
Machine Learning
Offline reinforcement learning (RL) has progressed with return-conditioned supervised learning (RCSL), but its lack of stitching ability remains a limitation. We introduce $Q$-Aided Conditional Supervised Learning (QCS), which effectively combines the stability of RCSL with the stitching capability of $Q$-functions. By analyzing $Q$-function over-generalization, which impairs stable stitching, QCS adaptively integrates $Q$-aid into RCSL's loss function based on trajectory return. Empirical results show that QCS significantly outperforms RCSL and value-based methods, consistently achieving or exceeding the maximum trajectory returns across diverse offline RL benchmarks.
title Adaptive $Q$-Aid for Conditional Supervised Learning in Offline Reinforcement Learning
topic Machine Learning
url https://arxiv.org/abs/2402.02017