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 2008.06199cs.LGstat.ML
收藏

Advertisement Agnostic Robust Deep强化学习

Adversary Agnostic Robust Deep Reinforcement Learning

Nanyang Technological University(南洋理工大学);Singapore Institute of Manufacturing Technology(新加坡制造技术研究院);Agency for Science, Technology and Research(新加坡科技研究局);Macquarie University(麦考瑞大学)

Xinghua Qu, Yew-Soon Ong, Abhishek Gupta, Zhu Sun

摘要:

深度强化学习(DRL)策略已被证明会被扰动所欺骗(例如,随机噪声或密集的对抗攻击)对在测试时出现但在训练期间未知的状态观测进行的。为了提高DRL策略的鲁棒性,以前的方法假设可以将对手的知识添加到训练过程中,以实现对这些扰动观测的相应泛化能力。然而,这样的假设不仅使鲁棒性改进更加昂贵,而且还可能使模型对其他类型的攻击不那么有效。相比之下,我们提出了一个对手不可知的强大的DRL范式,不需要从对手学习。为此,我们首先从理论上推导出,鲁棒性确实可以实现独立的对手的基础上的政策蒸馏设置。基于这一发现,我们提出了一种新的策略蒸馏损失,它有两个项:1)处方间隙最大化损失,旨在同时最大化教师策略选择的动作的可能性和剩余动作的熵; 2)相应的雅可比正则化损失,最小化相对于输入状态的梯度的大小。理论分析表明,我们的蒸馏损失保证增加处方差距和对抗鲁棒性。此外,五个Atari游戏的实验坚定地验证了我们的方法在提高对抗鲁棒性方面的优越性,与其他最先进的方法相比。

英文摘要:

Deep reinforcement learning (DRL) policies have been shown to be deceived by perturbations (e.g., random noise or intensional adversarial attacks) on state observations that appear at test time but are unknown during training. To increase the robustness of DRL policies, previous approaches assume that the knowledge of adversaries can be added into the training process to achieve the corresponding generalization ability on these perturbed observations. However, such an assumption not only makes the robustness improvement more expensive but may also leave a model less effective to other kinds of attacks in the wild. In contrast, we propose an adversary agnostic robust DRL paradigm that does not require learning from adversaries. To this end, we first theoretically derive that robustness could indeed be achieved independently of the adversaries based on a policy distillation setting. Motivated by this finding, we propose a new policy distillation loss with two terms: 1) a prescription gap maximization loss aiming at simultaneously maximizing the likelihood of the action selected by the teacher policy and the entropy over the remaining actions; 2) a corresponding Jacobian regularization loss that minimizes the magnitude of the gradient with respect to the input state. The theoretical analysis shows that our distillation loss guarantees to increase the prescription gap and the adversarial robustness. Furthermore, experiments on five Atari games firmly verify the superiority of our approach in terms of boosting adversarial robustness compared to other state-of-the-art methods.

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。

↑