Meta Reinforcement Learning (MRL) enables an agent to learn from a limited number of past trajectories and extrapolate to a new task. In this paper, we attempt to improve the robustness of MRL. We build upon model-agnostic meta-learning (MAML) and propose a novel method to generate adversarial samples for MRL by using Generative Adversarial Network (GAN). That allows us to enhance the robustness of MRL to adversal attacks by leveraging these attacks during meta training process.
翻译:元加强学习使代理商能够从有限的过去轨迹中学习,并推断出新的任务。在本文件中,我们试图提高MRL的稳健性。我们利用模型-不可知性元学习(MAML),提出了一种新颖的方法,利用基因反向网络(GAN)为MRL生成对抗性样本。这使我们能够通过在元培训过程中利用这些攻击,提高MRL对反向攻击的稳健性。