Auxiliary Adversarial

Auxiliary adversarial methods enhance the robustness of machine learning models, particularly in reinforcement learning and visual object tracking, by training against artificially generated adversarial examples. Current research focuses on developing algorithms that generate diverse and challenging adversarial attacks, often employing evolutionary strategies or multi-agent frameworks, and integrating these attacks into the training process of the main model. This approach aims to improve the generalization and resilience of learned policies to unexpected perturbations or noisy inputs, leading to more reliable and safe applications in areas like robotics and autonomous systems.

Papers

June 14, 2024

Robust Model-Based Reinforcement Learning with an Adversarial Auxiliary Model
Siemen Herremans, Ali Anwar, Siegfried Mercelis
Reinforcement Learning Markov Decision Process Robust Reinforcement Learning Auxiliary Adversarial

February 28, 2024

Enhancing Tracking Robustness with Auxiliary Adversarial Defense Networks
Zhewei Wu, Ruilong Yu, Qihe Liu, Shuying Cheng, Shilin Qiu, Shijie Zhou
Adversarial Attack Adversarial Training Adversarial Perturbation Adversarial Defense Robust Tracking Auxiliary Adversarial

May 10, 2023

Robust multi-agent coordination via evolutionary generation of auxiliary adversarial attackers
Lei Yuan, Zi-Qian Zhang, Ke Xue, Hao Yin, Feng Chen, Cong Guan, Li-He Li, Chao Qian, Yang Yu
Contrastive Learning Cooperative Multi Agent Reinforcement Learning Multi Agent Coordination Behavioral Diversity Evolutionary Generation Auxiliary Adversarial

May 9, 2023

Communication-Robust Multi-Agent Learning by Adaptable Auxiliary Multi-Agent Adversary Generation
Lei Yuan, Feng Chen, Zhongzhang Zhang, Yang Yu
Reinforcement Learning Multi Agent Adaptive Importance Cooperative Multi Agent Reinforcement Learning Communication Learning Auxiliary Adversarial

Auxiliary Adversarial

Papers

Robust Model-Based Reinforcement Learning with an Adversarial Auxiliary Model

Enhancing Tracking Robustness with Auxiliary Adversarial Defense Networks

Robust multi-agent coordination via evolutionary generation of auxiliary adversarial attackers

Communication-Robust Multi-Agent Learning by Adaptable Auxiliary Multi-Agent Adversary Generation