Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2405.14082 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866912083880181760 |
|---|---|
| author | Yeom, Junghyuk Jo, Yonghyeon Kim, Jungmo Lee, Sanghyeon Han, Seungyul |
| author_facet | Yeom, Junghyuk Jo, Yonghyeon Kim, Jungmo Lee, Sanghyeon Han, Seungyul |
| contents | Constraint-based offline reinforcement learning (RL) involves policy constraints or imposing penalties on the value function to mitigate overestimation errors caused by distributional shift. This paper focuses on a limitation in existing offline RL methods with penalized value function, indicating the potential for underestimation bias due to unnecessary bias introduced in the value function. To address this concern, we propose Exclusively Penalized Q-learning (EPQ), which reduces estimation bias in the value function by selectively penalizing states that are prone to inducing estimation errors. Numerical results show that our method significantly reduces underestimation bias and improves performance in various offline control tasks compared to other offline RL methods |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2405_14082 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Exclusively Penalized Q-learning for Offline Reinforcement Learning Yeom, Junghyuk Jo, Yonghyeon Kim, Jungmo Lee, Sanghyeon Han, Seungyul Machine Learning Artificial Intelligence Constraint-based offline reinforcement learning (RL) involves policy constraints or imposing penalties on the value function to mitigate overestimation errors caused by distributional shift. This paper focuses on a limitation in existing offline RL methods with penalized value function, indicating the potential for underestimation bias due to unnecessary bias introduced in the value function. To address this concern, we propose Exclusively Penalized Q-learning (EPQ), which reduces estimation bias in the value function by selectively penalizing states that are prone to inducing estimation errors. Numerical results show that our method significantly reduces underestimation bias and improves performance in various offline control tasks compared to other offline RL methods |
| title | Exclusively Penalized Q-learning for Offline Reinforcement Learning |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2405.14082 |