Model reusability in Reinforcement Learning
The VLDB Journal · 2025-07 · 34 (4)
DOI: 10.1007/s00778-025-00920-0 ↗Kurz gefasst
Fachartikel in der Zeitschrift The VLDB Journal, online erschienen am 12.05.2025. Verfasst von Sepideh Nikookar und 4 weiteren Autor:innen. Die Arbeit wurde bisher 10-mal zitiert (Crossref) und nennt 52 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Division of Information and Intelligent Systems, Office of Naval Research Global, Office of Naval Research.
Abstract (Originalsprache)
The ability to reuse trained models in Reinforcement Learning (RL) holds substantial practical value in particular for complex tasks. While model reusability is widely studied for supervised models in data management, to the best of our knowledge, this is the first ever principled study that is proposed for RL. To capture trained policies, we develop a framework based on an expressive and lossless graph data model that accommodates Temporal Difference Learning and Deep-RL based RL algorithms. Our framework is able to capture arbitrary reward functions that can be composed at inference time. The framework comes with theoretical guarantees and shows that it yields the same result as policies trained from scratch. We design a parameterized algorithm that strikes a balance between efficiency and quality w.r.t cumulative reward. Our experiments with two common RL tasks (query refinement and robot movement) corroborate our theory and show the effectiveness and efficiency of our algorithms.
Förderung
- Division of Information and Intelligent Systems
- Office of Naval Research Global
- Office of Naval Research
Zitieren
APA
Nikookar, S., Nia, S. N., Roy, S. B., Amer-Yahia, S., & Omidvar-Tehrani, B. (2025). Model reusability in Reinforcement Learning. The VLDB Journal, 34(4). https://doi.org/10.1007/s00778-025-00920-0
Vancouver
Nikookar S, Nia SN, Roy SB, Amer-Yahia S, Omidvar-Tehrani B. Model reusability in Reinforcement Learning. The VLDB Journal. 2025;34(4). doi:10.1007/s00778-025-00920-0
BibTeX
@article{Nikookar2025,
author = {Nikookar, Sepideh and Nia, Sohrab Namazi and Roy, Senjuti Basu and Amer-Yahia, Sihem and Omidvar-Tehrani, Behrooz},
title = {Model reusability in Reinforcement Learning},
journal = {The VLDB Journal},
year = {2025},
volume = {34},
number = {4},
doi = {10.1007/s00778-025-00920-0}
}
EUVIMED