Two-Step Offline Preference-Based Reinforcement Learning with Constrained Actions
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2023
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866914987494080512 |
|---|---|
| author | Xu, Yinglun Suresh, Tarun Gumaste, Rohan Zhu, David Li, Ruirui Wang, Zhengyang Jiang, Haoming Tang, Xianfeng Yin, Qingyu Cheng, Monica Xiao Zeng, Qi Zhang, Chao Singh, Gagandeep |
| author_facet | Xu, Yinglun Suresh, Tarun Gumaste, Rohan Zhu, David Li, Ruirui Wang, Zhengyang Jiang, Haoming Tang, Xianfeng Yin, Qingyu Cheng, Monica Xiao Zeng, Qi Zhang, Chao Singh, Gagandeep |
| contents | Preference-based reinforcement learning (PBRL) in the offline setting has succeeded greatly in industrial applications such as chatbots. A two-step learning framework where one applies a reinforcement learning step after a reward modeling step has been widely adopted for the problem. However, such a method faces challenges from the risk of reward hacking and the complexity of reinforcement learning. To overcome the challenge, our insight is that both challenges come from the state-actions not supported in the dataset. Such state-actions are unreliable and increase the complexity of the reinforcement learning problem at the second step. Based on the insight, we develop a novel two-step learning method called PRC: preference-based reinforcement learning with constrained actions. The high-level idea is to limit the reinforcement learning agent to optimize over a constrained action space that excludes the out-of-distribution state-actions. We empirically verify that our method has high learning efficiency on various datasets in robotic control environments. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2401_00330 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | Two-Step Offline Preference-Based Reinforcement Learning with Constrained Actions Xu, Yinglun Suresh, Tarun Gumaste, Rohan Zhu, David Li, Ruirui Wang, Zhengyang Jiang, Haoming Tang, Xianfeng Yin, Qingyu Cheng, Monica Xiao Zeng, Qi Zhang, Chao Singh, Gagandeep Machine Learning Artificial Intelligence Preference-based reinforcement learning (PBRL) in the offline setting has succeeded greatly in industrial applications such as chatbots. A two-step learning framework where one applies a reinforcement learning step after a reward modeling step has been widely adopted for the problem. However, such a method faces challenges from the risk of reward hacking and the complexity of reinforcement learning. To overcome the challenge, our insight is that both challenges come from the state-actions not supported in the dataset. Such state-actions are unreliable and increase the complexity of the reinforcement learning problem at the second step. Based on the insight, we develop a novel two-step learning method called PRC: preference-based reinforcement learning with constrained actions. The high-level idea is to limit the reinforcement learning agent to optimize over a constrained action space that excludes the out-of-distribution state-actions. We empirically verify that our method has high learning efficiency on various datasets in robotic control environments. |
| title | Two-Step Offline Preference-Based Reinforcement Learning with Constrained Actions |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2401.00330 |