Wasserstein Uncertainty

Wasserstein uncertainty focuses on developing robust machine learning models that are resilient to uncertainties in data distributions, often represented as Wasserstein balls around an empirical distribution. Current research emphasizes developing efficient algorithms for solving distributionally robust optimization problems, particularly using techniques like Wasserstein proximal gradient flows and incorporating Wasserstein distances into generative models for uncertainty quantification. This approach enhances model robustness against adversarial attacks, data scarcity, and domain shifts, with applications spanning fault diagnosis, reinforcement learning, and adversarial training. The resulting methods improve the generalization and reliability of machine learning models in various real-world scenarios.

Papers

October 4, 2024

A Federated Distributionally Robust Support Vector Machine with Mixture of Wasserstein Balls Ambiguity Set for Distributed Fault Diagnosis
Michael Ibrahim, Heraldo Rozas, Nagi Gebraeel, Weijun Xie
Mixture Component Support Vector Machine ADMM Algorithm Federated Machine Learning Wasserstein Uncertainty

July 18, 2024

Distributionally and Adversarially Robust Logistic Regression via Intersecting Wasserstein Balls
Aras Selvi, Eleonora Kreacic, Mohsen Ghassemi, Vamsi Potluru, Tucker Balch, Manuela Veloso
Adversarial Attack Native Robustness Empirical Risk Minimization Robust Optimization Distributional Assumption Robust Regression Sliced Wasserstein Wasserstein Uncertainty

July 16, 2024

Global Optimisation of Black-Box Functions with Generative Models in the Wasserstein Space
Tigran Ramazyan, Mikhail Hushchyn, Denis Derkach
Generative Model Global Optimization Gradient Free Black Box Function Wasserstein Space Wasserstein Uncertainty Generative Surrogate

June 24, 2024

Differentiable Distributionally Robust Optimization Layers
Xutao Ma, Chao Ning, Wenli Du
Robust Optimization Distributionally Robust Optimization Ambiguity Set Differentiable Surrogate Wasserstein Uncertainty

May 24, 2024

Score-based generative models are provably robust: an uncertainty quantification perspective
Nikiforos Mimikos-Stamatopoulos, Benjamin J. Zhang, Markos A. Katsoulakis
Uncertainty Quantification Score Based Generative Wasserstein Uncertainty Score Estimation Network

October 30, 2023

Flow-based Distributionally Robust Optimization
Chen Xu, Jonghyeok Lee, Xiuyuan Cheng, Yao Xie
Robust Optimization Worst Case Distributionally Robust Optimization Robust Algorithm Transport Map Wasserstein Uncertainty

March 5, 2023

Improved Sample Complexity Bounds for Distributionally Robust Reinforcement Learning
Zaiyan Xu, Kishan Panaganti, Dileep Kalathil
Robust Reinforcement Learning Distributionally Robust Sample Complexity Bound Uncertainty Set Wasserstein Uncertainty

September 30, 2022

Robust $Q$-learning Algorithm for Markov Decision Processes under Wasserstein Uncertainty
Ariel Neufeld, Julian Sester
Practical Algorithm Markov Decision Process Robust Version Distributional Robustness Stochastic Optimal Control Q$ Learning Algorithm Wasserstein Uncertainty Robust Markov Decision

July 11, 2022

Generalizing to Unseen Domains with Wasserstein Distributional Robustness under Limited Source Knowledge
Jingge Wang, Liyan Xie, Yao Xie, Shao-Lun Huang, Yang Li
Strong Generalization Domain Generalization Unseen Domain Target Domain Limited Information Wasserstein Distributionally Robust Wasserstein Uncertainty

January 22, 2022

The Many Faces of Adversarial Risk
Muni Sreenivas Pydi, Varun Jog
Adversarial Robustness Adversarial Risk Unbalanced Optimal Transport Wasserstein Uncertainty

Wasserstein Uncertainty

Papers

A Federated Distributionally Robust Support Vector Machine with Mixture of Wasserstein Balls Ambiguity Set for Distributed Fault Diagnosis

Distributionally and Adversarially Robust Logistic Regression via Intersecting Wasserstein Balls

Global Optimisation of Black-Box Functions with Generative Models in the Wasserstein Space

Differentiable Distributionally Robust Optimization Layers

Score-based generative models are provably robust: an uncertainty quantification perspective

Flow-based Distributionally Robust Optimization

Improved Sample Complexity Bounds for Distributionally Robust Reinforcement Learning

Robust $Q$-learning Algorithm for Markov Decision Processes under Wasserstein Uncertainty

Generalizing to Unseen Domains with Wasserstein Distributional Robustness under Limited Source Knowledge

The Many Faces of Adversarial Risk