Adaptive Adversarial Training for Meta Reinforcement Learning
This work addresses robustness issues in MRL for applications requiring security against adversarial threats, but it appears incremental as it builds upon existing MAML frameworks.
The paper tackles the problem of improving robustness in Meta Reinforcement Learning (MRL) by proposing a novel method that uses a Generative Adversarial Network (GAN) to generate adversarial samples, enhancing robustness to adversarial attacks during meta-training.
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.