Hierarchical Transformers are Efficient Meta-Reinforcement Learners
Fuente:
arXiv
Salvato in:
| Autori principali: | , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866909100109987840 |
|---|---|
| author | Shala, Gresa Biedenkapp, André Grabocka, Josif |
| author_facet | Shala, Gresa Biedenkapp, André Grabocka, Josif |
| contents | We introduce Hierarchical Transformers for Meta-Reinforcement Learning (HTrMRL), a powerful online meta-reinforcement learning approach. HTrMRL aims to address the challenge of enabling reinforcement learning agents to perform effectively in previously unseen tasks. We demonstrate how past episodes serve as a rich source of information, which our model effectively distills and applies to new contexts. Our learned algorithm is capable of outperforming the previous state-of-the-art and provides more efficient meta-training while significantly improving generalization capabilities. Experimental results, obtained across various simulated tasks of the Meta-World Benchmark, indicate a significant improvement in learning efficiency and adaptability compared to the state-of-the-art on a variety of tasks. Our approach not only enhances the agent's ability to generalize from limited data but also paves the way for more robust and versatile AI systems. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2402_06402 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Hierarchical Transformers are Efficient Meta-Reinforcement Learners Shala, Gresa Biedenkapp, André Grabocka, Josif Machine Learning Artificial Intelligence We introduce Hierarchical Transformers for Meta-Reinforcement Learning (HTrMRL), a powerful online meta-reinforcement learning approach. HTrMRL aims to address the challenge of enabling reinforcement learning agents to perform effectively in previously unseen tasks. We demonstrate how past episodes serve as a rich source of information, which our model effectively distills and applies to new contexts. Our learned algorithm is capable of outperforming the previous state-of-the-art and provides more efficient meta-training while significantly improving generalization capabilities. Experimental results, obtained across various simulated tasks of the Meta-World Benchmark, indicate a significant improvement in learning efficiency and adaptability compared to the state-of-the-art on a variety of tasks. Our approach not only enhances the agent's ability to generalize from limited data but also paves the way for more robust and versatile AI systems. |
| title | Hierarchical Transformers are Efficient Meta-Reinforcement Learners |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2402.06402 |