arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2407.07082cs.LGcs.AI
收藏

学习优化能否降低强化学习的难度?

Can Learned Optimization Make Reinforcement Learning Less Difficult?

University of Oxford(牛津大学)

Alexander David Goldie, Chris Lu, Matthew Thomas Jackson, Shimon Whiteson, Jakob Nicolaus Foerster

AI总结:

针对强化学习的非平稳性、可塑性损失及探索问题,提出OPEN方法,通过元学习参考现有解决方案的更新规则,在多样场景中灵活元学习并利用随机性探索,实验显示其优于或等于传统优化器且泛化性强。

摘要:

尽管强化学习(RL)在现实世界决策中具有巨大潜力,但它面临诸多独特困难,往往需要针对性考量:高度非平稳性、严重的可塑性损失,以及需通过探索避免过早收敛到局部最优并最大化回报。本文探讨学习优化能否帮助克服这些问题。我们提出的方法——面向可塑性、探索与非平稳性的学习优化(OPEN),元学习一种更新规则,其输入特征和输出结构参考了此前针对这些困难提出的解决方案。我们证明该参数化足够灵活,可支持多样学习场景下的元学习,包括利用随机性进行探索的能力。实验表明,在单一及小型环境集上进行元训练后,OPEN优于或等于传统优化器;此外,OPEN在多种环境和智能体架构中展现出强大的泛化特性。

英文摘要:

While reinforcement learning (RL) holds great potential for decision making in the real world, it suffers from a number of unique difficulties which often need specific consideration. In particular: it is highly non-stationary; suffers from high degrees of plasticity loss; and requires exploration to prevent premature convergence to local optima and maximize return. In this paper, we consider whether learned optimization can help overcome these problems. Our method, Learned Optimization for Plasticity, Exploration and Non-stationarity (OPEN), meta-learns an update rule whose input features and output structure are informed by previously proposed solutions to these difficulties. We show that our parameterization is flexible enough to enable meta-learning in diverse learning contexts, including the ability to use stochasticity for exploration. Our experiments demonstrate that when meta-trained on single and small sets of environments, OPEN outperforms or equals traditionally used optimizers. Furthermore, OPEN shows strong generalization characteristics across a range of environments and agent architectures.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑