arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2106.01404cs.LGcs.AI
收藏

变分赋权作为基于目标的强化学习的表示学习

Variational Empowerment as Representation Learning for Goal-Based Reinforcement Learning

University of Michigan(密歇根大学);Stanford University(斯坦福大学);LG AI Research(LG人工智能研究院);Google Research(谷歌研究院);University of California, Berkeley(加州大学伯克利分校)

Jongwook Choi, Archit Sharma, Honglak Lee, Sergey Levine, Shixiang Shane Gu

摘要:

学习达到目标状态和学习不同的技能,通过互信息(MI)最大化已被提出作为原则性框架的自我监督强化学习,允许代理获得广泛适用的多任务策略,最小的奖励工程。从一个简单的观察,标准的目标条件RL(GCRL)封装的优化目标的变分授权,我们讨论如何GCRL和MI为基础的RL可以推广到一个单一的家庭的方法,我们命名为变分GCRL(VGCRL),解释变分MI最大化,或变分授权,作为表示学习方法,获得功能感知的状态表示,达到目标。这种新颖的视角使我们能够:(1)推导出简单但未经探索的GCRL变体,以研究如何增加小的表示容量就可以扩展其功能;(2)通过修改潜在维度和应用谱归一化,研究递归函数容量和平滑度如何决定所发现的技能或潜在目标的质量;(3)将后见之明经验回放(HER)等技术从GCRL应用到基于MI的RL;(4)提出一种新的评价指标--潜在目标达成(LGR),用于比较具有不同潜在维度和潜在目标参数化选择的授权算法。通过原则性的数学推导和细致的实验研究,我们的工作奠定了一个新的基础,从评估,分析和发展表示学习技术在基于目标的强化学习。

英文摘要:

Learning to reach goal states and learning diverse skills through mutual information (MI) maximization have been proposed as principled frameworks for self-supervised reinforcement learning, allowing agents to acquire broadly applicable multitask policies with minimal reward engineering. Starting from a simple observation that the standard goal-conditioned RL (GCRL) is encapsulated by the optimization objective of variational empowerment, we discuss how GCRL and MI-based RL can be generalized into a single family of methods, which we name variational GCRL (VGCRL), interpreting variational MI maximization, or variational empowerment, as representation learning methods that acquire functionally-aware state representations for goal reaching. This novel perspective allows us to: (1) derive simple but unexplored variants of GCRL to study how adding small representation capacity can already expand its capabilities; (2) investigate how discriminator function capacity and smoothness determine the quality of discovered skills, or latent goals, through modifying latent dimensionality and applying spectral normalization; (3) adapt techniques such as hindsight experience replay (HER) from GCRL to MI-based RL; and lastly, (4) propose a novel evaluation metric, named latent goal reaching (LGR), for comparing empowerment algorithms with different choices of latent dimensionality and discriminator parameterization. Through principled mathematical derivations and careful experimental studies, our work lays a novel foundation from which to evaluate, analyze, and develop representation learning techniques in goal-based RL.

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。

↑