Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2405.09999 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917822804787200 |
|---|---|
| author | Naik, Abhishek Wan, Yi Tomar, Manan Sutton, Richard S. |
| author_facet | Naik, Abhishek Wan, Yi Tomar, Manan Sutton, Richard S. |
| contents | We show that discounted methods for solving continuing reinforcement learning problems can perform significantly better if they center their rewards by subtracting out the rewards' empirical average. The improvement is substantial at commonly used discount factors and increases further as the discount factor approaches one. In addition, we show that if a problem's rewards are shifted by a constant, then standard methods perform much worse, whereas methods with reward centering are unaffected. Estimating the average reward is straightforward in the on-policy setting; we propose a slightly more sophisticated method for the off-policy setting. Reward centering is a general idea, so we expect almost every reinforcement-learning algorithm to benefit by the addition of reward centering. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2405_09999 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Reward Centering Naik, Abhishek Wan, Yi Tomar, Manan Sutton, Richard S. Machine Learning Artificial Intelligence We show that discounted methods for solving continuing reinforcement learning problems can perform significantly better if they center their rewards by subtracting out the rewards' empirical average. The improvement is substantial at commonly used discount factors and increases further as the discount factor approaches one. In addition, we show that if a problem's rewards are shifted by a constant, then standard methods perform much worse, whereas methods with reward centering are unaffected. Estimating the average reward is straightforward in the on-policy setting; we propose a slightly more sophisticated method for the off-policy setting. Reward centering is a general idea, so we expect almost every reinforcement-learning algorithm to benefit by the addition of reward centering. |
| title | Reward Centering |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2405.09999 |