Distributed robust Gaussian Process regression

Research output: Journal contributionsJournal articlesResearchpeer-review

Standard

Distributed robust Gaussian Process regression. / Mair, Sebastian; Brefeld, Ulf.
In: Knowledge and Information Systems, Vol. 55, No. 2, 01.05.2018, p. 415-435.

Research output: Journal contributionsJournal articlesResearchpeer-review

Harvard

APA

Vancouver

Mair S, Brefeld U. Distributed robust Gaussian Process regression. Knowledge and Information Systems. 2018 May 1;55(2):415-435. Epub 2017 Jul 19. doi: 10.1007/s10115-017-1084-7

Bibtex

@article{e7e0e3faacc7473b99d76b870a3b77d6,
title = "Distributed robust Gaussian Process regression",
abstract = "We study distributed and robust Gaussian Processes where robustness is introduced by a Gaussian Process prior on the function values combined with a Student-t likelihood. The posterior distribution is approximated by a Laplace Approximation, and together with concepts from Bayesian Committee Machines, we efficiently distribute the computations and render robust GPs on huge data sets feasible. We provide a detailed derivation and report on empirical results. Our findings on real and artificial data show that our approach outperforms existing baselines in the presence of outliers by using all available data.",
keywords = "Business informatics, Distributed computation, Gaussian process regression, Laplace Approximation, Robust regression, Student-t likelihood",
author = "Sebastian Mair and Ulf Brefeld",
year = "2018",
month = may,
day = "1",
doi = "10.1007/s10115-017-1084-7",
language = "English",
volume = "55",
pages = "415--435",
journal = "Knowledge and Information Systems",
issn = "0219-1377",
publisher = "Springer UK",
number = "2",

}

RIS

TY - JOUR

T1 - Distributed robust Gaussian Process regression

AU - Mair, Sebastian

AU - Brefeld, Ulf

PY - 2018/5/1

Y1 - 2018/5/1

N2 - We study distributed and robust Gaussian Processes where robustness is introduced by a Gaussian Process prior on the function values combined with a Student-t likelihood. The posterior distribution is approximated by a Laplace Approximation, and together with concepts from Bayesian Committee Machines, we efficiently distribute the computations and render robust GPs on huge data sets feasible. We provide a detailed derivation and report on empirical results. Our findings on real and artificial data show that our approach outperforms existing baselines in the presence of outliers by using all available data.

AB - We study distributed and robust Gaussian Processes where robustness is introduced by a Gaussian Process prior on the function values combined with a Student-t likelihood. The posterior distribution is approximated by a Laplace Approximation, and together with concepts from Bayesian Committee Machines, we efficiently distribute the computations and render robust GPs on huge data sets feasible. We provide a detailed derivation and report on empirical results. Our findings on real and artificial data show that our approach outperforms existing baselines in the presence of outliers by using all available data.

KW - Business informatics

KW - Distributed computation

KW - Gaussian process regression

KW - Laplace Approximation

KW - Robust regression

KW - Student-t likelihood

UR - http://www.scopus.com/inward/record.url?scp=85025086950&partnerID=8YFLogxK

U2 - 10.1007/s10115-017-1084-7

DO - 10.1007/s10115-017-1084-7

M3 - Journal articles

VL - 55

SP - 415

EP - 435

JO - Knowledge and Information Systems

JF - Knowledge and Information Systems

SN - 0219-1377

IS - 2

ER -