arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2009.09052cs.LGcs.CRstat.ML
收藏

基于PAC和后悔保证的私有强化学习

Private Reinforcement Learning with PAC and Regret Guarantees

University of Minnesota(明尼苏达大学);DeepMind;Microsoft Research(微软研究院);Carnegie Mellon University(卡内基梅隆大学)

Giuseppe Vietri, Borja Balle, Akshay Krishnamurthy, Zhiwei Steven Wu

摘要:

受个性化医疗等用户信息固有敏感的高风险决策领域的激励,我们设计了用于情景强化学习(RL)的隐私保护探索策略。我们首先使用联合差分隐私(JDP)的概念提供一个有意义的隐私公式--差分隐私的一个强变体,用于每个用户接收他们自己的输出集的设置(例如,政策建议)。然后,我们开发了一个私人的基于乐观主义的学习算法,同时实现强大的PAC和遗憾的界限,并享有JDP保证。我们的算法只支付一个适度的隐私成本的探索:在比较非私人的界限,隐私参数只出现在较低的顺序。最后,我们提出了样本的复杂性和遗憾的强化学习服从JDP的下限。

英文摘要:

Motivated by high-stakes decision-making domains like personalized medicine where user information is inherently sensitive, we design privacy preserving exploration policies for episodic reinforcement learning (RL). We first provide a meaningful privacy formulation using the notion of joint differential privacy (JDP)--a strong variant of differential privacy for settings where each user receives their own sets of output (e.g., policy recommendations). We then develop a private optimism-based learning algorithm that simultaneously achieves strong PAC and regret bounds, and enjoys a JDP guarantee. Our algorithm only pays for a moderate privacy cost on exploration: in comparison to the non-private bounds, the privacy parameter only appears in lower-order terms. Finally, we present lower bounds on sample complexity and regret for reinforcement learning subject to JDP.

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。

↑