REVIEW 3 cited by
Online Robustness Training for Deep Reinforcement Learning
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
In deep reinforcement learning (RL), adversarial attacks can trick an agent into unwanted states and disrupt training. We propose a system called Robust Student-DQN (RS-DQN), which permits online robustness training alongside Q networks, while preserving competitive performance. We show that RS-DQN can be combined with (i) state-of-the-art adversarial training and (ii) provably robust training to obtain an agent that is resilient to strong attacks during training and evaluation.
Forward citations
Cited by 3 Pith papers
-
Action Robust Reinforcement Learning via Optimal Adversary Aware Policy Optimization
OA-PI trains RL policies against the worst-case bounded action perturbation through a modified Bellman operator, and its TD3 and PPO variants improve robustness and nominal performance on Mujoco and Box2d tasks.
-
Towards Robust Deep Reinforcement Learning against Environmental State Perturbation
Environmental perturbations to the initial state sharply reduce the rewards of PPO-trained Overcooked agents, and the proposed BAT defense, supervised kickstarting followed by adversarial fine-tuning, restores robustn...
-
Advancing Robustness in Deep Reinforcement Learning with an Ensemble Defense Approach
Averaging three observation filters (random noise, autoencoder, PCA) before action selection substantially improves a Highway-env DQN's reward and collision rate under FGSM attacks.
Discussion (0). Continue with ORCID to comment.