Criteria of efficiency for set-valued classification

Vladimir Vovk; Ilia Nouretdinov; Valentina Fedorova; Ivan Petej; Alexander Gammerman

doi:10.1007/s10472-017-9540-3

Criteria of efficiency for set-valued classification

Vladimir Vovk, Ilia Nouretdinov, Valentina Fedorova, Ivan Petej, Alexander Gammerman

Research output: Contribution to journal › Article › peer-review

114 Downloads (Pure)

Abstract

We study optimal conformity measures for various criteria of efficiency of set-valued classification in an idealised setting. This leads to an important class of criteria of efficiency that we call probabilistic and argue for; it turns out that the most standard criteria of efficiency used in literature on conformal prediction are not probabilistic unless the problem of classification is binary. We consider both unconditional and label-conditional conformal prediction.

Original language	English
Pages (from-to)	21-46
Number of pages	26
Journal	Annals of Mathematics and Artificial Intelligence
Volume	81
Early online date	14 Mar 2017
DOIs	https://doi.org/10.1007/s10472-017-9540-3
Publication status	Published - Oct 2017

Access to Document

10.1007/s10472-017-9540-3Licence: CC BY

Accepted ManuscriptAccepted author manuscript, 959 KBLicence: CC BY

Exascale Compound Activity Prediction Engine
Gammerman, A. (PI), Vovk, V. (CoI) & Luo, Z. (CoI)
European Union
1/09/15 → 31/08/18
Project: Research
Mining the Network Behaviour of Bots
Cavallaro, L. (PI), Gammerman, A. (CoI), Vovk, V. (CoI), Shanahan, H. (CoI) & Luo, Z. (CoI)
Eng & Phys Sci Res Council EPSRC
16/06/13 → 17/04/17
Project: Research

Cite this

@article{6712a49ed4364a0997aedc28223d68de,

title = "Criteria of efficiency for set-valued classification",

abstract = "We study optimal conformity measures for various criteria of efficiency of set-valued classification in an idealised setting. This leads to an important class of criteria of efficiency that we call probabilistic and argue for; it turns out that the most standard criteria of efficiency used in literature on conformal prediction are not probabilistic unless the problem of classification is binary. We consider both unconditional and label-conditional conformal prediction.",

author = "Vladimir Vovk and Ilia Nouretdinov and Valentina Fedorova and Ivan Petej and Alexander Gammerman",

year = "2017",

month = oct,

doi = "10.1007/s10472-017-9540-3",

language = "English",

volume = "81",

pages = "21--46",

journal = "Annals of Mathematics and Artificial Intelligence",

issn = "1012-2443",

publisher = "Springer Netherlands",

}

TY - JOUR

T1 - Criteria of efficiency for set-valued classification

AU - Vovk, Vladimir

AU - Nouretdinov, Ilia

AU - Fedorova, Valentina

AU - Petej, Ivan

AU - Gammerman, Alexander

PY - 2017/10

Y1 - 2017/10

N2 - We study optimal conformity measures for various criteria of efficiency of set-valued classification in an idealised setting. This leads to an important class of criteria of efficiency that we call probabilistic and argue for; it turns out that the most standard criteria of efficiency used in literature on conformal prediction are not probabilistic unless the problem of classification is binary. We consider both unconditional and label-conditional conformal prediction.

AB - We study optimal conformity measures for various criteria of efficiency of set-valued classification in an idealised setting. This leads to an important class of criteria of efficiency that we call probabilistic and argue for; it turns out that the most standard criteria of efficiency used in literature on conformal prediction are not probabilistic unless the problem of classification is binary. We consider both unconditional and label-conditional conformal prediction.

U2 - 10.1007/s10472-017-9540-3

DO - 10.1007/s10472-017-9540-3

M3 - Article

SN - 1012-2443

VL - 81

SP - 21

EP - 46

JO - Annals of Mathematics and Artificial Intelligence

JF - Annals of Mathematics and Artificial Intelligence

ER -

Criteria of efficiency for set-valued classification

Abstract

Access to Document

Projects

Exascale Compound Activity Prediction Engine

Mining the Network Behaviour of Bots

Cite this