AI Paper Atlas

Research map / Adversarial robustness and security / Certified and empirical robustness

Robust Deep Reinforcement Learning through Adversarial Loss

Tuomas Oikarinen, Wang Zhang, Alexandre Megretski, Luca Daniel, Tsui-Wei Weng

NeurIPS 2021 · Adversarial robustness and security › Certified and empirical robustness · approach: Provable robust learning guarantees · 15 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use Approach 2 overlap loss instead of Approach 1 upper-bound loss (ablation)RADIAL-DQN · Atari RoadRunner PGD 1/255reward: 4240 -> 44445Improved
use Approach 2 instead of Approach 1 (ablation)RADIAL-DQN · Atari Pong PGD 5/255reward: 1.1 -> 21.0Improved
use Approach 2 instead of Approach 1 (ablation)RADIAL-DQN · Atari BankHeist PGD 5/255reward: 1063.5 -> 1182.5Improved
margin coefficient c=0.5 instead of 0.25 (ablation)RADIAL-DQN · Atari RoadRunner PGD 5/255reward: 11220 -> 23820Improved
margin coefficient c=0.5 instead of 0.75 (ablation)RADIAL-DQN · Atari RoadRunner PGD 5/255reward: 4870 -> 23820Improved
symmetric overlap loss instead of default (ablation)RADIAL-DQN · Atari RoadRunner/BankHeist PGD 5/255reward: RoadRunner 23820 -> 21385; BankHeist 1182.5 -> 1305Mixed
CROWN-IBP bounds instead of IBP (ablation)RADIAL-DQN · Atari PGD 5/255reward: RoadRunner 23820 -> 18315; BankHeist 1182.5 -> 1328.5Hurt
apply RADIAL training (Approach 2) to DQN versus standard DQN (ablation)DQN · Atari RoadRunner PGD 1/255reward: 0.0 -> 44445Improved
RADIAL-DQN versus SA-DQNDQN · Atari RoadRunner PGD 5/255reward: 3350 -> 23820Improved
RADIAL-PPO versus PPOPPO · ProcGen Fruitbot Eval PGD 5/255reward: 8.51 -> 25.81Improved
RADIAL-PPO versus SA-PPOPPO · MuJoCo Walker2D MAD 0.15reward: 3652.1 -> 3895.3Improved
RADIAL-PPO versus SA-PPOPPO · MuJoCo HalfCheetah MAD 0.15reward: 4173.1 -> 4008.5Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy