Deep neural policies have recently been installed in a diverse range of settings, from biotechnology to automated financial systems. However, the utilization of deep neural networks to approximate the value function leads to concerns on the decision boundary stability, in particular, with regard to the sensitivity of policy decision making to indiscernible, non-robust features due to highly non-convex and complex deep neural manifolds. These concerns constitute an obstruction to understanding the reasoning made by deep neural policies, and their foundational limitations. Hence, it is crucial to develop techniques that aim to understand the sensitivities in the learnt representations of neural network policies. To achieve this we introduce a theoretically founded method that provides a systematic analysis of the unstable directions in the deep neural policy decision boundary across both time and space. Through experiments in the Arcade Learning Environment (ALE), we demonstrate the effectiveness of our technique for identifying correlated directions of instability, and for measuring how sample shifts remold the set of sensitive directions in the neural policy landscape. Most importantly, we demonstrate that state-of-the-art robust training techniques yield learning of disjoint unstable directions, with dramatically larger oscillations over time, when compared to standard training. We believe our results reveal the fundamental properties of the decision process made by reinforcement learning policies, and can help in constructing reliable and robust deep neural policies.

Learning from raw high dimensional data via interaction with a given environment has been effectively achieved through the utilization of deep neural networks. Yet the observed degradation in policy performance caused by imperceptible worst-case policy dependent translations along high sensitivity directions (i.e. adversarial perturbations) raises concerns on the robustness of deep reinforcement learning policies. In our paper, we show that these high sensitivity directions do not lie only along particular worst-case directions, but rather are more abundant in the deep neural policy landscape and can be found via more natural means in a black-box setting. Furthermore, we show that vanilla training techniques intriguingly result in learning more robust policies compared to the policies learnt via the state-of-the-art adversarial training techniques. We believe our work lays out intriguing properties of the deep reinforcement learning policy manifold and our results can help to build robust and generalizable deep reinforcement learning policies.

Reinforcement learning policies based on deep neural networks are vulnerable to imperceptible adversarial perturbations to their inputs, in much the same way as neural network image classifiers. Recent work has proposed several methods to improve the robustness of deep reinforcement learning agents to adversarial perturbations based on training in the presence of these imperceptible perturbations (i.e. adversarial training). In this paper, we study the effects of adversarial training on the neural policy learned by the agent. In particular, we follow two distinct parallel approaches to investigate the outcomes of adversarial training on deep neural policies based on worst-case distributional shift and feature sensitivity. For the first approach, we compare the Fourier spectrum of minimal perturbations computed for both adversarially trained and vanilla trained neural policies. Via experiments in the OpenAI Atari environments we show that minimal perturbations computed for adversarially trained policies are more focused on lower frequencies in the Fourier domain, indicating a higher sensitivity of these policies to low frequency perturbations. For the second approach, we propose a novel method to measure the feature sensitivities of deep neural policies and we compare these feature sensitivity differences in state-of-the-art adversarially trained deep neural policies and vanilla trained deep neural policies. We believe our results can be an initial step towards understanding the relationship between adversarial training and different notions of robustness for neural policies.

Policy optimization is an effective reinforcement learning approach to solve continuous control tasks. Recent achievements have shown that alternating on-line and off-line optimization is a successful choice for efficient trajectory reuse. However, deciding when to stop optimizing and collect new trajectories is non-trivial as it requires to account for the variance of the objective function estimate. In this paper, we propose a novel model-free policy search algorithm, POIS, applicable in both control-based and parameter-based settings. We first derive a high-confidence bound for importance sampling estimation and then we define a surrogate objective function which is optimized off-line using a batch of trajectories. Finally, the algorithm is tested on a selection of continuous control tasks, with both linear and deep policies, and compared with the state-of-the-art policy optimization methods.