Robust Risk

Robust risk research focuses on developing machine learning models and statistical methods that are resilient to uncertainties and adversarial attacks, aiming to minimize worst-case performance degradation. Current efforts concentrate on developing computationally efficient algorithms for distributionally robust optimization, often employing techniques like Wasserstein distance or Sinkhorn divergence to quantify uncertainty, and exploring the benefits of incorporating regularization or local queries to enhance robustness. This work is crucial for deploying reliable machine learning systems in safety-critical applications and improving the generalizability and trustworthiness of models across diverse real-world scenarios.

Papers

August 19, 2024

Regularization for Adversarial Robust Learning
Jie Wang, Rui Gao, Yao Xie
Adversarial Attack Empirical Risk Minimization Gradient Norm Divergence Regularization Robust Risk

March 21, 2024

Non-Convex Robust Hypothesis Testing using Sinkhorn Uncertainty Sets
Jie Wang, Rui Gao, Yao Xie
Hypothesis Testing Uncertainty Set Distributional Uncertainty Robust Risk

February 26, 2023

Efficient physics-informed neural networks using hash encoding
Xinquan Huang, Tariq Alkhalifah
High Efficiency Physic Informed Neural Network Numerical Method Finite Difference Hash Code Hash Encoding Robust Risk

January 3, 2023

Risk-Averse MDPs under Reward Ambiguity
Haolin Ruan, Zhi Chen, Chin Pang Ho
Markov Decision Process Reward Ambiguity Robust Markov Decision Process Robust Risk

October 12, 2022

When are Local Queries Useful for Robust Learning?
Pascale Gourdeau, Varun Kanade, Marta Kwiatkowska, James Worrell
Empirical Risk Minimization Robust Learning Robust Risk

Robust Risk

Papers

Regularization for Adversarial Robust Learning

Non-Convex Robust Hypothesis Testing using Sinkhorn Uncertainty Sets

Efficient physics-informed neural networks using hash encoding

Risk-Averse MDPs under Reward Ambiguity

When are Local Queries Useful for Robust Learning?