An open problem on strongly consistent learning of the best prediction for Gaussian processes

L. Györfi, Alessio Sancetta

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

For Gaussian process, we present an open problem whether or not there is a data driven predictor of the conditional expectation of the current value given the past such that the difference between the predictor and the conditional expectation tends to zero almost surely for all stationary, ergodic, Gaussian processes. We show some related negative and positive findings.

Original languageEnglish
Title of host publicationTopics in Nonparametric Statistics - Proceedings of the 1st Conference of the International Society for Nonparametric Statistics
PublisherSpringer New York LLC
Pages115-136
Number of pages22
Volume74
ISBN (Print)9781493905683
DOIs
Publication statusPublished - 2014
Event1st Conference of the International Society of Nonparametric Statistics, ISNPS 2012 - Chalkidiki, Greece
Duration: Jun 15 2012Jun 19 2012

Other

Other1st Conference of the International Society of Nonparametric Statistics, ISNPS 2012
CountryGreece
CityChalkidiki
Period6/15/126/19/12

Fingerprint

Conditional Expectation
Gaussian Process
Predictors
Open Problems
Prediction
Data-driven
Tend
Zero
Learning

Keywords

  • Conditional expectation
  • Data driven predictor
  • Ergodic process
  • Gaussian time series
  • Linear regression
  • Strong consistency

ASJC Scopus subject areas

  • Mathematics(all)

Cite this

Györfi, L., & Sancetta, A. (2014). An open problem on strongly consistent learning of the best prediction for Gaussian processes. In Topics in Nonparametric Statistics - Proceedings of the 1st Conference of the International Society for Nonparametric Statistics (Vol. 74, pp. 115-136). Springer New York LLC. https://doi.org/10.1007/978-1-4939-0569-0_12

An open problem on strongly consistent learning of the best prediction for Gaussian processes. / Györfi, L.; Sancetta, Alessio.

Topics in Nonparametric Statistics - Proceedings of the 1st Conference of the International Society for Nonparametric Statistics. Vol. 74 Springer New York LLC, 2014. p. 115-136.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Györfi, L & Sancetta, A 2014, An open problem on strongly consistent learning of the best prediction for Gaussian processes. in Topics in Nonparametric Statistics - Proceedings of the 1st Conference of the International Society for Nonparametric Statistics. vol. 74, Springer New York LLC, pp. 115-136, 1st Conference of the International Society of Nonparametric Statistics, ISNPS 2012, Chalkidiki, Greece, 6/15/12. https://doi.org/10.1007/978-1-4939-0569-0_12
Györfi L, Sancetta A. An open problem on strongly consistent learning of the best prediction for Gaussian processes. In Topics in Nonparametric Statistics - Proceedings of the 1st Conference of the International Society for Nonparametric Statistics. Vol. 74. Springer New York LLC. 2014. p. 115-136 https://doi.org/10.1007/978-1-4939-0569-0_12
Györfi, L. ; Sancetta, Alessio. / An open problem on strongly consistent learning of the best prediction for Gaussian processes. Topics in Nonparametric Statistics - Proceedings of the 1st Conference of the International Society for Nonparametric Statistics. Vol. 74 Springer New York LLC, 2014. pp. 115-136
@inproceedings{b6ee6b4b0da34e6dbde23ebd6a7a2ac2,
title = "An open problem on strongly consistent learning of the best prediction for Gaussian processes",
abstract = "For Gaussian process, we present an open problem whether or not there is a data driven predictor of the conditional expectation of the current value given the past such that the difference between the predictor and the conditional expectation tends to zero almost surely for all stationary, ergodic, Gaussian processes. We show some related negative and positive findings.",
keywords = "Conditional expectation, Data driven predictor, Ergodic process, Gaussian time series, Linear regression, Strong consistency",
author = "L. Gy{\"o}rfi and Alessio Sancetta",
year = "2014",
doi = "10.1007/978-1-4939-0569-0_12",
language = "English",
isbn = "9781493905683",
volume = "74",
pages = "115--136",
booktitle = "Topics in Nonparametric Statistics - Proceedings of the 1st Conference of the International Society for Nonparametric Statistics",
publisher = "Springer New York LLC",

}

TY - GEN

T1 - An open problem on strongly consistent learning of the best prediction for Gaussian processes

AU - Györfi, L.

AU - Sancetta, Alessio

PY - 2014

Y1 - 2014

N2 - For Gaussian process, we present an open problem whether or not there is a data driven predictor of the conditional expectation of the current value given the past such that the difference between the predictor and the conditional expectation tends to zero almost surely for all stationary, ergodic, Gaussian processes. We show some related negative and positive findings.

AB - For Gaussian process, we present an open problem whether or not there is a data driven predictor of the conditional expectation of the current value given the past such that the difference between the predictor and the conditional expectation tends to zero almost surely for all stationary, ergodic, Gaussian processes. We show some related negative and positive findings.

KW - Conditional expectation

KW - Data driven predictor

KW - Ergodic process

KW - Gaussian time series

KW - Linear regression

KW - Strong consistency

UR - http://www.scopus.com/inward/record.url?scp=84920024788&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84920024788&partnerID=8YFLogxK

U2 - 10.1007/978-1-4939-0569-0_12

DO - 10.1007/978-1-4939-0569-0_12

M3 - Conference contribution

SN - 9781493905683

VL - 74

SP - 115

EP - 136

BT - Topics in Nonparametric Statistics - Proceedings of the 1st Conference of the International Society for Nonparametric Statistics

PB - Springer New York LLC

ER -