PAC-Bayesian Soft Actor-Critic Learning
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2023
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866917688950915072 |
|---|---|
| author | Tasdighi, Bahareh Akgül, Abdullah Haussmann, Manuel Brink, Kenny Kazimirzak Kandemir, Melih |
| author_facet | Tasdighi, Bahareh Akgül, Abdullah Haussmann, Manuel Brink, Kenny Kazimirzak Kandemir, Melih |
| contents | Actor-critic algorithms address the dual goals of reinforcement learning (RL), policy evaluation and improvement via two separate function approximators. The practicality of this approach comes at the expense of training instability, caused mainly by the destructive effect of the approximation errors of the critic on the actor. We tackle this bottleneck by employing an existing Probably Approximately Correct (PAC) Bayesian bound for the first time as the critic training objective of the Soft Actor-Critic (SAC) algorithm. We further demonstrate that online learning performance improves significantly when a stochastic actor explores multiple futures by critic-guided random search. We observe our resulting algorithm to compare favorably against the state-of-the-art SAC implementation on multiple classical control and locomotion tasks in terms of both sample efficiency and regret. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2301_12776 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | PAC-Bayesian Soft Actor-Critic Learning Tasdighi, Bahareh Akgül, Abdullah Haussmann, Manuel Brink, Kenny Kazimirzak Kandemir, Melih Machine Learning Actor-critic algorithms address the dual goals of reinforcement learning (RL), policy evaluation and improvement via two separate function approximators. The practicality of this approach comes at the expense of training instability, caused mainly by the destructive effect of the approximation errors of the critic on the actor. We tackle this bottleneck by employing an existing Probably Approximately Correct (PAC) Bayesian bound for the first time as the critic training objective of the Soft Actor-Critic (SAC) algorithm. We further demonstrate that online learning performance improves significantly when a stochastic actor explores multiple futures by critic-guided random search. We observe our resulting algorithm to compare favorably against the state-of-the-art SAC implementation on multiple classical control and locomotion tasks in terms of both sample efficiency and regret. |
| title | PAC-Bayesian Soft Actor-Critic Learning |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2301.12776 |