An Advantage-based Optimization Method for Reinforcement Learning in Large Action Space

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Lin, Hai, Huang, Cheng, Chen, Zhihong
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866929635049078784
author Lin, Hai
Huang, Cheng
Chen, Zhihong
author_facet Lin, Hai
Huang, Cheng
Chen, Zhihong
contents Reinforcement learning tasks in real-world scenarios often involve large, high-dimensional action spaces, leading to challenges such as convergence difficulties, instability, and high computational complexity. It is widely acknowledged that traditional value-based reinforcement learning algorithms struggle to address these issues effectively. A prevalent approach involves generating independent sub-actions within each dimension of the action space. However, this method introduces bias, hindering the learning of optimal policies. In this paper, we propose an advantage-based optimization method and an algorithm named Advantage Branching Dueling Q-network (ABQ). ABQ incorporates a baseline mechanism to tune the action value of each dimension, leveraging the advantage relationship across different sub-actions. With this approach, the learned policy can be optimized for each dimension. Empirical results demonstrate that ABQ outperforms BDQ, achieving 3%, 171%, and 84% more cumulative rewards in HalfCheetah, Ant, and Humanoid environments, respectively. Furthermore, ABQ exhibits competitive performance when compared against two continuous action benchmark algorithms, DDPG and TD3.
format Preprint
id arxiv_https___arxiv_org_abs_2412_12605
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle An Advantage-based Optimization Method for Reinforcement Learning in Large Action Space
Lin, Hai
Huang, Cheng
Chen, Zhihong
Machine Learning
Artificial Intelligence
Reinforcement learning tasks in real-world scenarios often involve large, high-dimensional action spaces, leading to challenges such as convergence difficulties, instability, and high computational complexity. It is widely acknowledged that traditional value-based reinforcement learning algorithms struggle to address these issues effectively. A prevalent approach involves generating independent sub-actions within each dimension of the action space. However, this method introduces bias, hindering the learning of optimal policies. In this paper, we propose an advantage-based optimization method and an algorithm named Advantage Branching Dueling Q-network (ABQ). ABQ incorporates a baseline mechanism to tune the action value of each dimension, leveraging the advantage relationship across different sub-actions. With this approach, the learned policy can be optimized for each dimension. Empirical results demonstrate that ABQ outperforms BDQ, achieving 3%, 171%, and 84% more cumulative rewards in HalfCheetah, Ant, and Humanoid environments, respectively. Furthermore, ABQ exhibits competitive performance when compared against two continuous action benchmark algorithms, DDPG and TD3.
title An Advantage-based Optimization Method for Reinforcement Learning in Large Action Space
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2412.12605