Conformal calibration

Vladimir Vovk; Ivan Petej; Paolo Toccaceli; Alex Gammerman; Ernst Ahlberg; Lars Carlsson

Conformal calibration

Vladimir Vovk, Ivan Petej, Paolo Toccaceli, Alex Gammerman, Ernst Ahlberg, Lars Carlsson

Research output: Chapter in Book/Report/Conference proceeding › Chapter

Abstract

Most existing examples of full conformal predictive systems, split conformal predictive systems, and cross-conformal predictive systems impose severe restrictions on the adaptation of predictive distributions to the test object at hand. In this paper we develop split conformal predictive systems that are fully adaptive. Our method consists in calibrating existing predictive systems; the input predictive system is not supposed to satisfy any properties of validity, whereas the output predictive system is guaranteed to be calibrated in probability.

Original language	English
Title of host publication	Proceedings of Machine Learning Research
Editors	Alex Gammerman, Vladimir Vovk, Zhiyuan Luo, Evgueni Smirnov, Giovanni Cherubin
Pages	84-99
Number of pages	16
Volume	128
Publication status	Published - Sept 2020

Keywords

conformal prediction
cross-conformal prediction
predictive distributions

Cite this

@inbook{cebcc86f39844bc0ad644944c1fb4e4e,

title = "Conformal calibration",

abstract = "Most existing examples of full conformal predictive systems, split conformal predictive systems, and cross-conformal predictive systems impose severe restrictions on the adaptation of predictive distributions to the test object at hand. In this paper we develop split conformal predictive systems that are fully adaptive. Our method consists in calibrating existing predictive systems; the input predictive system is not supposed to satisfy any properties of validity, whereas the output predictive system is guaranteed to be calibrated in probability.",

keywords = "conformal prediction, cross-conformal prediction, predictive distributions",

author = "Vladimir Vovk and Ivan Petej and Paolo Toccaceli and Alex Gammerman and Ernst Ahlberg and Lars Carlsson",

year = "2020",

month = sep,

language = "English",

volume = "128",

pages = "84--99",

editor = "Alex Gammerman and Vladimir Vovk and Zhiyuan Luo and Evgueni Smirnov and Giovanni Cherubin",

booktitle = "Proceedings of Machine Learning Research",

}

TY - CHAP

T1 - Conformal calibration

AU - Vovk, Vladimir

AU - Petej, Ivan

AU - Toccaceli, Paolo

AU - Gammerman, Alex

AU - Ahlberg, Ernst

AU - Carlsson, Lars

PY - 2020/9

Y1 - 2020/9

N2 - Most existing examples of full conformal predictive systems, split conformal predictive systems, and cross-conformal predictive systems impose severe restrictions on the adaptation of predictive distributions to the test object at hand. In this paper we develop split conformal predictive systems that are fully adaptive. Our method consists in calibrating existing predictive systems; the input predictive system is not supposed to satisfy any properties of validity, whereas the output predictive system is guaranteed to be calibrated in probability.

AB - Most existing examples of full conformal predictive systems, split conformal predictive systems, and cross-conformal predictive systems impose severe restrictions on the adaptation of predictive distributions to the test object at hand. In this paper we develop split conformal predictive systems that are fully adaptive. Our method consists in calibrating existing predictive systems; the input predictive system is not supposed to satisfy any properties of validity, whereas the output predictive system is guaranteed to be calibrated in probability.

KW - conformal prediction

KW - cross-conformal prediction

KW - predictive distributions

M3 - Chapter

VL - 128

SP - 84

EP - 99

BT - Proceedings of Machine Learning Research

A2 - Gammerman, Alex

A2 - Vovk, Vladimir

A2 - Luo, Zhiyuan

A2 - Smirnov, Evgueni

A2 - Cherubin, Giovanni

ER -