Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2405.00410 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913894168002560 |
|---|---|
| author | Shi, Yucheng Lynch, David Agapitos, Alexandros |
| author_facet | Shi, Yucheng Lynch, David Agapitos, Alexandros |
| contents | In Multi-objective Reinforcement Learning (MORL) agents are tasked with optimising decision-making behaviours that trade-off between multiple, possibly conflicting, objectives. MORL based on decomposition is a family of solution methods that employ a number of utility functions to decompose the multi-objective problem into individual single-objective problems solved simultaneously in order to approximate a Pareto front of policies. We focus on the case of linear utility functions parametrised by weight vectors w. We introduce a method based on Upper Confidence Bound to efficiently search for the most promising weight vectors during different stages of the learning process, with the aim of maximising the hypervolume of the resulting Pareto front. The proposed method demonstrates consistency and strong performance across various MORL baselines on Mujoco benchmark problems. The code is released in: https://github.com/SYCAMORE-1/ucb-MOPPO |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2405_00410 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | UCB-driven Utility Function Search for Multi-objective Reinforcement Learning Shi, Yucheng Lynch, David Agapitos, Alexandros Machine Learning In Multi-objective Reinforcement Learning (MORL) agents are tasked with optimising decision-making behaviours that trade-off between multiple, possibly conflicting, objectives. MORL based on decomposition is a family of solution methods that employ a number of utility functions to decompose the multi-objective problem into individual single-objective problems solved simultaneously in order to approximate a Pareto front of policies. We focus on the case of linear utility functions parametrised by weight vectors w. We introduce a method based on Upper Confidence Bound to efficiently search for the most promising weight vectors during different stages of the learning process, with the aim of maximising the hypervolume of the resulting Pareto front. The proposed method demonstrates consistency and strong performance across various MORL baselines on Mujoco benchmark problems. The code is released in: https://github.com/SYCAMORE-1/ucb-MOPPO |
| title | UCB-driven Utility Function Search for Multi-objective Reinforcement Learning |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2405.00410 |