Adaptive conformal classification with noisy labels

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Sesia, Matteo, Wang, Y. X. Rachel, Tong, Xin
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929251265019904
author Sesia, Matteo
Wang, Y. X. Rachel
Tong, Xin
author_facet Sesia, Matteo
Wang, Y. X. Rachel
Tong, Xin
contents This paper develops novel conformal prediction methods for classification tasks that can automatically adapt to random label contamination in the calibration sample, leading to more informative prediction sets with stronger coverage guarantees compared to state-of-the-art approaches. This is made possible by a precise characterization of the effective coverage inflation (or deflation) suffered by standard conformal inferences in the presence of label contamination, which is then made actionable through new calibration algorithms. Our solution is flexible and can leverage different modeling assumptions about the label contamination process, while requiring no knowledge of the underlying data distribution or of the inner workings of the machine-learning classifier. The advantages of the proposed methods are demonstrated through extensive simulations and an application to object classification with the CIFAR-10H image data set.
format Preprint
id arxiv_https___arxiv_org_abs_2309_05092
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Adaptive conformal classification with noisy labels
Sesia, Matteo
Wang, Y. X. Rachel
Tong, Xin
Methodology
Machine Learning
Statistics Theory
This paper develops novel conformal prediction methods for classification tasks that can automatically adapt to random label contamination in the calibration sample, leading to more informative prediction sets with stronger coverage guarantees compared to state-of-the-art approaches. This is made possible by a precise characterization of the effective coverage inflation (or deflation) suffered by standard conformal inferences in the presence of label contamination, which is then made actionable through new calibration algorithms. Our solution is flexible and can leverage different modeling assumptions about the label contamination process, while requiring no knowledge of the underlying data distribution or of the inner workings of the machine-learning classifier. The advantages of the proposed methods are demonstrated through extensive simulations and an application to object classification with the CIFAR-10H image data set.
title Adaptive conformal classification with noisy labels
topic Methodology
Machine Learning
Statistics Theory
url https://arxiv.org/abs/2309.05092