Distribution-free uncertainty quantification for classification under label shift

2021/03/04 by Aleksandr Podkopaev, Aaditya Ramdas, Podkopaev, Aleksandr +1 · 16 citations
Computer Science · Mathematics · #Adversarial Robustness in Machine Learning #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Machine Learning and Algorithms #Machine Learning and Data Classification #cs.LG #stat.ML

paper · pdf · doi:10.48550/arxiv.2103.03323

openalex publication_date 2021/03/04 · arxiv created 2021/07/07 · arxiv updated 2021/07/08 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28

Abstract

Trustworthy deployment of ML models requires a proper measure of uncertainty, especially in safety-critical applications. We focus on uncertainty quantification (UQ) for classification problems via two avenues -- prediction sets using conformal prediction and calibration of probabilistic predictors by post-hoc binning -- since these possess distribution-free guarantees for i.i.d. data. Two common ways of generalizing beyond the i.i.d. setting include handling covariate and label shift. Within the context of distribution-free UQ, the former has already received attention, but not the latter. It is known that label shift hurts prediction, and we first argue that it also hurts UQ, by showing degradation in coverage and calibration. Piggybacking on recent progress in addressing label shift (for better prediction), we examine the right way to achieve UQ by reweighting the aforementioned conformal and calibration procedures whenever some unlabeled data from the target distribution is available. We examine these techniques theoretically in a distribution-free framework and demonstrate their excellent practical performance.

Citations

• Adjusting the Outputs of a Classifier to New aPrioriProbabilities: A Simple Procedure
• Least Ambiguous Set-Valued Classifiers with Bounded Error Levels
• Regularized Learning for Domain Adaptation under Label Shifts
• Evaluating model calibration in classification
• Beyond temperature scaling: Obtaining well-calibrated multiclass probabilities with Dirichlet calibration

Cited by

• Beyond Aggregate Risk: Role-Stratified Conformal Risk Control for LLM Tool Calls
• Weighted Conformal Prediction Provides Adaptive and Valid Mask-Conditional Coverage for General Missing Data Mechanisms
• Elements of Conformal Prediction for Statisticians
• Distribution-Free Prediction Sets for Regression under Target Shift
• Protected Probabilistic Classification Library
• Conformal Prediction Meets Long-tail Classification
• Conformal Unlearning: A New Paradigm for Unlearning in Conformal Predictors
• Non-exchangeable Conformal Prediction with Optimal Transport: Tackling Distribution Shifts with Unlabeled Data
• Conformal Prediction for Long-Tailed Classification
• Multi-model Online Conformal Prediction with Graph-Structured Feedback
• On Continuous Monitoring of Risk Violations under Unknown Shift
• CP²: Leveraging Geometry for Conformal Prediction via Canonicalization
• WATCH: Adaptive Monitoring for AI Deployments via Weighted-Conformal Martingales
• Backward Conformal Prediction
• Synthetic-Powered Predictive Inference
• Conformalized Large Language Models under Configuration Shift
• A Deployment Audit of Release-Side Risk in Conformal Triage under Prevalence Shift

Related

• Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift
• A comparative study of conformal prediction methods for valid uncertainty quantification in machine learning
• Adapting Prediction Sets to Distribution Shifts Without Labels
• Distribution-Free, Risk-Controlling Prediction Sets
• Conformal Prediction for Federated Uncertainty Quantification Under Label Shift