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 2102.06587cs.LGcs.AI
收藏

具有损坏奖励的干扰强化学习代理

Disturbing Reinforcement Learning Agents with Corrupted Rewards

Universidad Carlos III de Madrid(马德里卡洛斯三世大学)

Rubén Majadas, Javier García, Fernando Fernández

摘要:

强化学习(RL)算法最近在解决复杂游戏(如Atari或Starcraft)方面取得了成功,并在现实世界的应用中产生了巨大的影响,如网络安全或自动驾驶。在缺点方面,最近的工作已经表明RL算法的性能如何在奖励函数的软变化的影响下下降。然而,很少有工作已经做了如何敏感这些干扰取决于攻击的侵略性和学习探索策略。在本文中,我们建议填补这一空白,在文献中分析不同的攻击策略的奖励扰动的基础上的影响,并研究在学习器中的效果取决于其探索策略。为了解释所有的行为,我们选择了一个子类的MDP:情节,随机目标唯一的奖励MDP,特别是,一个可理解的网格域作为基准。在这个领域中,我们证明了平滑地制作对抗性奖励能够误导学习者,并且使用低探索概率值,学习到的策略对腐败奖励更鲁棒。最后,在所提出的学习场景中,出现了一个违反直觉的结果:在每个学习片段中攻击是成本最低的攻击策略。

英文摘要:

Reinforcement Learning (RL) algorithms have led to recent successes in solving complex games, such as Atari or Starcraft, and to a huge impact in real-world applications, such as cybersecurity or autonomous driving. In the side of the drawbacks, recent works have shown how the performance of RL algorithms decreases under the influence of soft changes in the reward function. However, little work has been done about how sensitive these disturbances are depending on the aggressiveness of the attack and the learning exploration strategy. In this paper, we propose to fill this gap in the literature analyzing the effects of different attack strategies based on reward perturbations, and studying the effect in the learner depending on its exploration strategy. In order to explain all the behaviors, we choose a sub-class of MDPs: episodic, stochastic goal-only-rewards MDPs, and in particular, an intelligible grid domain as a benchmark. In this domain, we demonstrate that smoothly crafting adversarial rewards are able to mislead the learner, and that using low exploration probability values, the policy learned is more robust to corrupt rewards. Finally, in the proposed learning scenario, a counterintuitive result arises: attacking at each learning episode is the lowest cost attack strategy.

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。

↑