Conditional Shift

Conditional shift, a type of distribution shift in machine learning, refers to discrepancies in the conditional probability of a label given an input between training and testing data. Current research focuses on mitigating this shift's negative impact on model performance across various domains, including graph neural networks, time-series forecasting, and image classification, employing techniques like adversarial training, conformal prediction, and causal structure analysis within models such as DeepONets and GNNs. Addressing conditional shift is crucial for improving the reliability and generalizability of machine learning models in real-world applications where data distributions inevitably differ between training and deployment environments. This research directly impacts the robustness and trustworthiness of AI systems across numerous fields.

Papers

October 9, 2024

LLM Embeddings Improve Test-time Adaptation to Tabular $Y|X$-Shifts
Yibo Zeng, Jiashuo Liu, Henry Lam, Hongseok Namkoong
Tabular Data Test Time Table Semantics Tabular Datasets LLM Embeddings Informative Representation Conditional Shift Pseudo Target

May 20, 2024

Conditional Shift-Robust Conformal Prediction for Graph Neural Network
S. Akansha
Graph Neural Network Conformal Prediction Graph Structured Data Conditional Reasoning GNN Prediction Robust Conformal Prediction Conditional Shift

September 25, 2023

Hierarchical Imitation Learning for Stochastic Environments
Maximilian Igl, Punit Shah, Paul Mougin, Sirish Srinivasan, Tarun Gupta, Brandyn White, Kyriacos Shiarlis, Shimon Whiteson
Imitation Learning Stochastic Environment Behavior Model Conditional Shift

August 29, 2023

Biquality Learning: a Framework to Design Algorithms Dealing with Closed-Set Distribution Shifts
Pierre Nodet, Vincent Lemaire, Alexis Bondu, Antoine Cornuéjols
New Framework Weak Supervision Algorithm Design Dataset Shift Conditional Shift

July 11, 2023

Rethinking Distribution Shifts: Empirical Analysis and Inductive Modeling for Tabular Data
Jiashuo Liu, Tianyu Wang, Peng Cui, Hongseok Namkoong
Distribution Shift Robust Optimization Community Need Covariate Shift Tabular Datasets Modeling Language Hand Drawn Illustration Conditional Shift

June 5, 2023

September 17, 2022

Mitigating Both Covariate and Conditional Shift for Domain Generalization
Jianxin Lin, Yongqiang Tang, Junping Wang, Wensheng Zhang
Domain Generalization Covariate Shift Target Domain Relevant Covariates Visual Alignment Conditional Shift

April 20, 2022

Deep transfer operator learning for partial differential equations under conditional shift
Somdatta Goswami, Katiana Kontolati, Michael D. Shields, George Em Karniadakis
Partial Differential Equation Task Specific Deep Operator Network Functional Regression Conditional Shift

November 5, 2021

Transferable Time-Series Forecasting under Causal Conditional Shift
Zijian Li, Ruichu Cai, Tom Z. J Fu, Zhifeng Hao, Kun Zhang
Semi Supervised Domain Adaptation Time Series Domain Adaptation Conditional Shift Transferable Time Series