Home  | Publications | HJO+25

Conformal Prediction Without Nonconformity Scores

MCML Authors

Link to Profile Eyke Hüllermeier PI Matchmaking

Eyke Hüllermeier

Prof. Dr.

Principal Investigator

Artificial Intelligence and Machine Learning

Abstract

Conformal prediction (CP) is an uncertainty quantification framework that allows for constructing<br>statistically valid prediction sets. Key to the construction of these sets is the notion of nonconformity function, which assigns a real-valued score to individual data points: only those (hypothetical) data points contribute to a prediction set that sufficiently conform to the data. The point of departure of this work is the observation that CP predictions are invariant against (strictly) monotone transformations of a nonconformity function. In other words, it is only the ordering of the scores that matters, not their quantitative values. Consequently, instead of scoring individual data points, a conformal predictor only needs to be able to compare pairs of data points, deciding which of them is the more conforming one. This suggests an interesting connection between CP and preference learning, in particular learning-to-rank methods, and makes CP amenable to training data in the form of (qualitative) preferences. Elaborating on<br>this connection, we propose methods for learning (latent) nonconformity functions from data of that<br>kind and show their usefulness in real-world classification tasks.

inproceedings

UAI 2025

41st Conference on Uncertainty in Artificial Intelligence. Rio de Janeiro, Brazil, Jul 21-25, 2025.
Conference logo
A Conference

Authors

J. Hanselle • A. Javanmardi • T. Oberkofler • Y. Sale • E. Hüllermeier

Links

URL

Research Area

 A3 | Computational Models

BibTeXKey: HJO+25

Back to Top