Adversarial Robust Deep Reinforcement Learning Requires Redefining Robustness

Authors: Ezgi Korkmaz

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We run multiple experiments in the Arcade Learning Environment (ALE)
Researcher Affiliation Academia The paper only lists the author name 'Ezgi Korkmaz' without any institutional affiliation or email domain.
Pseudocode Yes Algorithm 1: Probing Neural Manifold with High-sensitivity Directions within Perceptual Similarity
Open Source Code No The paper does not provide any statement or link indicating the availability of its own source code.
Open Datasets Yes The Arcade Learning Environment (Bellemare et al. 2013).
Dataset Splits No The paper states that results are from '10 independent runs' but does not specify explicit training, validation, or test dataset splits.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory specifications) used for running the experiments.
Software Dependencies No The paper mentions several algorithms and frameworks like 'Deep Q-Network', 'prioritized experience replay', 'SA-DDQN', 'RADIAL', and 'Open AI Gym', but it does not specify version numbers for any of them.
Experiment Setup No The paper describes the general training methods and perturbation parameters (e.g., brightness/contrast values, rotation degrees) but does not provide specific hyperparameters for the deep reinforcement learning policy training itself, such as learning rate, batch size, or optimizer settings.