Covariate Shift Generalization

Covariate shift generalization focuses on improving the performance of machine learning models when the distribution of input features (covariates) differs between training and testing data. Current research emphasizes developing model-agnostic methods, such as multicalibration and bisimulation-based approaches, and algorithms that leverage sample reweighting and sparsity constraints to address this challenge, particularly in reinforcement learning and high-dimensional settings. These advancements aim to enhance the robustness and reliability of machine learning models in real-world applications where data distributions inevitably shift, leading to more dependable predictions and decisions across diverse contexts.

Papers

October 15, 2024

FOOGD: Federated Collaboration for Both Out-of-distribution Generalization and Detection
Xinting Liao, Weiming Liu, Pengyang Zhou, Fengyuan Yu, Jiahe Xu, Jun Wang, Wenjie Wang, Chaochao Chen, Xiaolin Zheng
Strong Generalization Data Detection Distribution Data Semantic Change Detection Covariate Shift Generalization

June 2, 2024

Bridging Multicalibration and Out-of-distribution Generalization Beyond Covariate Shift
Jiayun Wu, Jiashuo Liu, Peng Cui, Zhiwei Steven Wu
Native Robustness Distribution Generalization Robust Learning Covariate Shift Covariate Shift Generalization

June 7, 2023

Generalization Across Observation Shifts in Reinforcement Learning
Anuj Mahajan, Amy Zhang
Reinforcement Learning Reinforcement Learning Agent Bisimulation Metric Covariate Shift Generalization

December 2, 2022

Stable Learning via Sparse Variable Independence
Han Yu, Peng Cui, Yue He, Zheyan Shen, Yong Lin, Renzhe Xu, Xingxuan Zhang
Stable Learning Sample Reweighting Covariate Shift Generalization

November 3, 2021

A Theoretical Analysis on Independence-driven Importance Weighting for Covariate-shift Generalization
Renzhe Xu, Xingxuan Zhang, Zheyan Shen, Tong Zhang, Peng Cui
Loss Function Theoretical Analysis Covariate Shift Stable Learning Importance Weighting Covariate Shift Generalization

Covariate Shift Generalization

Papers

FOOGD: Federated Collaboration for Both Out-of-distribution Generalization and Detection

Bridging Multicalibration and Out-of-distribution Generalization Beyond Covariate Shift

Generalization Across Observation Shifts in Reinforcement Learning

Stable Learning via Sparse Variable Independence

A Theoretical Analysis on Independence-driven Importance Weighting for Covariate-shift Generalization