Conformal e-prediction

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autore principale: Vovk, Vladimir
Natura: Preprint
Pubblicazione: 2020
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910948133961728
author Vovk, Vladimir
author_facet Vovk, Vladimir
contents This paper discusses a counterpart of conformal prediction for e-values, conformal e-prediction. Conformal e-prediction is conceptually simpler and had been developed in the 1990s as a precursor of conformal prediction. When conformal prediction emerged as result of replacing e-values by p-values, it seemed to have important advantages over conformal e-prediction without obvious disadvantages. This paper re-examines relations between conformal prediction and conformal e-prediction systematically from a modern perspective. Conformal e-prediction has advantages of its own, such as the ease of designing conditional conformal e-predictors and the guaranteed validity of cross-conformal e-predictors (whereas for cross-conformal predictors validity is only an empirical fact and can be broken with excessive randomization). Even where conformal prediction has clear advantages, conformal e-prediction can often emulate those advantages, more or less successfully.
format Preprint
id arxiv_https___arxiv_org_abs_2001_05989
institution arXiv
publishDate 2020
record_format arxiv
spellingShingle Conformal e-prediction
Vovk, Vladimir
Machine Learning
68T05 (Primary) 68Q32, 62G15 (Secondary)
This paper discusses a counterpart of conformal prediction for e-values, conformal e-prediction. Conformal e-prediction is conceptually simpler and had been developed in the 1990s as a precursor of conformal prediction. When conformal prediction emerged as result of replacing e-values by p-values, it seemed to have important advantages over conformal e-prediction without obvious disadvantages. This paper re-examines relations between conformal prediction and conformal e-prediction systematically from a modern perspective. Conformal e-prediction has advantages of its own, such as the ease of designing conditional conformal e-predictors and the guaranteed validity of cross-conformal e-predictors (whereas for cross-conformal predictors validity is only an empirical fact and can be broken with excessive randomization). Even where conformal prediction has clear advantages, conformal e-prediction can often emulate those advantages, more or less successfully.
title Conformal e-prediction
topic Machine Learning
68T05 (Primary) 68Q32, 62G15 (Secondary)
url https://arxiv.org/abs/2001.05989