A Critical Perspective on Finite Sample Conformal Prediction Theory in Medical Applications

2025/12/09 by Kladny, Klaus-Rudolf, Schölkopf, Bernhard, Koch, Lisa +2
Computer Science · #Adversarial Robustness in Machine Learning #Artificial Intelligence (cs.AI) #Explainable Artificial Intelligence (XAI) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning in Healthcare #Methodology (stat.ME)

paper · doi:10.48550/arxiv.2512.14727

openalex publication_date 2025/12/09 · openalex created_date 2025/12/19 · openalex updated_date 2026/07/28

Abstract

Machine learning (ML) is transforming healthcare, but safe clinical decisions demand reliable uncertainty estimates that standard ML models fail to provide. Conformal prediction (CP) is a popular tool that allows users to turn heuristic uncertainty estimates into uncertainty estimates with statistical guarantees. CP works by converting predictions of a ML model, together with a calibration sample, into prediction sets that are guaranteed to contain the true label with any desired probability. An often cited advantage is that CP theory holds for calibration samples of arbitrary size, suggesting that uncertainty estimates with practically meaningful statistical guarantees can be achieved even if only small calibration sets are available. We question this promise by showing that, although the statistical guarantees hold for calibration sets of arbitrary size, the practical utility of these guarantees does highly depend on the size of the calibration set. This observation is relevant in medical domains because data is often scarce and obtaining large calibration sets is therefore infeasible. We corroborate our critique in an empirical demonstration on a medical image classification task.

Citations

• Pitfalls of Conformal Predictions for Medical Image Classification
• Adaptable Cardiovascular Disease Risk Prediction from Heterogeneous Data using Large Language Models
• Conformal Generative Modeling with Improved Sample Efficiency through Sequential Greedy Filtering
• Conformal Performance Range Prediction for Segmentation Output Quality Control
• Conformal prediction beyond exchangeability
• Learn then Test: Calibrating Predictive Algorithms to Achieve Risk Control
• Fair Conformal Predictors for Applications in Medical Imaging
• A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantification
• Selection and Aggregation of Conformal Prediction Sets
• Distribution-Free, Risk-Controlling Prediction Sets
• Distribution-free, Risk-controlling Prediction Sets
• PAC Confidence Sets for Deep Neural Networks via Calibrated Prediction
• Calibration: the Achilles heel of predictive analytics
• A Conformal Prediction Approach to Explore Functional Data
• Conditional validity of inductive conformal predictors

Related

• A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantification
• RR-CP: Reliable-Region-Based Conformal Prediction for Trustworthy Medical Image Classification
• Pitfalls of Conformal Predictions for Medical Image Classification
• Generalization and Informativeness of Conformal Prediction
• Reliable Statistical Guarantees for Conformal Predictors with Small Datasets