Transductive Conformal Prediction

Vladimir Vovk

doi:10.1142/S0218213015600015

Transductive Conformal Prediction

Vladimir Vovk

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

50 Downloads (Pure)

Abstract

This paper discusses transductive versions of conformal predictors and inductive conformal predictors. The transductive versions are computationally inefficient for long test sequences, but it turns out that apparently crude "Bonferroni predictors" are about as good in their informational efficiency and vastly superior in computational efficiency. The paper explores transductive predictors both theoretically and experimentally, in the latter case using the standard USPS data set of handwritten digits.

Original language	English
Article number	1560001
Journal	International Journal on Artificial Intelligence Tools
Volume	24
Issue number	6
DOIs	https://doi.org/10.1142/S0218213015600015
Publication status	Published - 21 Dec 2015

Keywords

Conformal predictors; inductive conformal predictors; transduction; Bonferroni adjustment.

Access to Document

10.1142/S0218213015600015

Accepted ManuscriptAccepted author manuscript, 386 KB

http://search.crossref.org/?q=10.1142%2FS0218213015600015

Cite this

@article{a931aa449a654fc7b6a48f0edae3cdf9,

title = "Transductive Conformal Prediction",

abstract = "This paper discusses transductive versions of conformal predictors and inductive conformal predictors. The transductive versions are computationally inefficient for long test sequences, but it turns out that apparently crude {"}Bonferroni predictors{"} are about as good in their informational efficiency and vastly superior in computational efficiency. The paper explores transductive predictors both theoretically and experimentally, in the latter case using the standard USPS data set of handwritten digits.",

keywords = "Conformal predictors; inductive conformal predictors; transduction; Bonferroni adjustment.",

author = "Vladimir Vovk",

year = "2015",

month = dec,

day = "21",

doi = "10.1142/S0218213015600015",

language = "English",

volume = "24",

journal = "International Journal on Artificial Intelligence Tools",

issn = "0218-2130",

publisher = "World Scientific Publishing Co. Pte Ltd",

number = "6",

}

TY - JOUR

T1 - Transductive Conformal Prediction

AU - Vovk, Vladimir

PY - 2015/12/21

Y1 - 2015/12/21

N2 - This paper discusses transductive versions of conformal predictors and inductive conformal predictors. The transductive versions are computationally inefficient for long test sequences, but it turns out that apparently crude "Bonferroni predictors" are about as good in their informational efficiency and vastly superior in computational efficiency. The paper explores transductive predictors both theoretically and experimentally, in the latter case using the standard USPS data set of handwritten digits.

AB - This paper discusses transductive versions of conformal predictors and inductive conformal predictors. The transductive versions are computationally inefficient for long test sequences, but it turns out that apparently crude "Bonferroni predictors" are about as good in their informational efficiency and vastly superior in computational efficiency. The paper explores transductive predictors both theoretically and experimentally, in the latter case using the standard USPS data set of handwritten digits.

KW - Conformal predictors; inductive conformal predictors; transduction; Bonferroni adjustment.

U2 - 10.1142/S0218213015600015

DO - 10.1142/S0218213015600015

M3 - Article

SN - 0218-2130

VL - 24

JO - International Journal on Artificial Intelligence Tools

JF - International Journal on Artificial Intelligence Tools

IS - 6

M1 - 1560001

ER -