PAC-Bayes Control: Learning Policies that Provably Generalize to Novel Environments
PAC-Bayes 控制:学习能够证明在新环境中泛化的能力的策略
机构 * Department of Mechanical and Aerospace Engineering(1,2 机械与航空航天工程系) ; Department of Computer Science Princeton University(3 计算机科学系 纽约大学普林斯顿分校)
AI总结 本文提出了一种基于PAC-Bayes框架的机器人策略学习方法,通过在新环境中泛化能力的理论分析,为机器人系统提供强泛化保证。
Comments Extended version of paper presented at the 2018 Conference on Robot Learning (CoRL)