Locally Smoothed Gaussian Process Regression

Davit Gogolashvili; Bogdan Kozyrskiy; Maurizio Filippone

doi:10.1016/j.procs.2022.09.330

Locally Smoothed Gaussian Process Regression

Davit Gogolashvili, Bogdan Kozyrskiy, Maurizio Filippone

Statistics

Research output: Contribution to conference › Paper › peer-review

6 Scopus citations

Abstract

We develop a novel framework to accelerate Gaussian process regression (GPR). In particular, we consider localization kernels at each data point to down-weigh the contributions from other data points that are far away, and we derive the GPR model stemming from the application of such localization operation. Through a set of experiments, we demonstrate the competitive performance of the proposed approach compared to full GPR, other localized models, and deep Gaussian processes. Crucially, these performances are obtained with considerable speedups compared to standard global GPR due to the sparsification effect of the Gram matrix induced by the localization operation.

Original language	English (US)
Pages	2717-2726
Number of pages	10
DOIs	https://doi.org/10.1016/j.procs.2022.09.330
State	Published - 2022
Event	26th International Conference on Knowledge-Based and Intelligent Information and Engineering Systems, KES 2022 - Verona, Italy Duration: Sep 7 2022 → Sep 9 2022

Conference

Conference	26th International Conference on Knowledge-Based and Intelligent Information and Engineering Systems, KES 2022
Country/Territory	Italy
City	Verona
Period	09/7/22 → 09/9/22

Keywords

Gaussian processes
kernel smoothing
local regression

ASJC Scopus subject areas

General Computer Science

Access to Document

10.1016/j.procs.2022.09.330

Cite this

@conference{0d830986fcc2470fab9056c3350725b1,

title = "Locally Smoothed Gaussian Process Regression",

abstract = "We develop a novel framework to accelerate Gaussian process regression (GPR). In particular, we consider localization kernels at each data point to down-weigh the contributions from other data points that are far away, and we derive the GPR model stemming from the application of such localization operation. Through a set of experiments, we demonstrate the competitive performance of the proposed approach compared to full GPR, other localized models, and deep Gaussian processes. Crucially, these performances are obtained with considerable speedups compared to standard global GPR due to the sparsification effect of the Gram matrix induced by the localization operation.",

keywords = "Gaussian processes, kernel smoothing, local regression",

author = "Davit Gogolashvili and Bogdan Kozyrskiy and Maurizio Filippone",

note = "Publisher Copyright: {\textcopyright} 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of the scientific committee of the 26th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems (KES 2022); 26th International Conference on Knowledge-Based and Intelligent Information and Engineering Systems, KES 2022 ; Conference date: 07-09-2022 Through 09-09-2022",

year = "2022",

doi = "10.1016/j.procs.2022.09.330",

language = "English (US)",

pages = "2717--2726",

}

TY - CONF

T1 - Locally Smoothed Gaussian Process Regression

AU - Gogolashvili, Davit

AU - Kozyrskiy, Bogdan

AU - Filippone, Maurizio

N1 - Publisher Copyright: © 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of the scientific committee of the 26th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems (KES 2022)

PY - 2022

Y1 - 2022

N2 - We develop a novel framework to accelerate Gaussian process regression (GPR). In particular, we consider localization kernels at each data point to down-weigh the contributions from other data points that are far away, and we derive the GPR model stemming from the application of such localization operation. Through a set of experiments, we demonstrate the competitive performance of the proposed approach compared to full GPR, other localized models, and deep Gaussian processes. Crucially, these performances are obtained with considerable speedups compared to standard global GPR due to the sparsification effect of the Gram matrix induced by the localization operation.

AB - We develop a novel framework to accelerate Gaussian process regression (GPR). In particular, we consider localization kernels at each data point to down-weigh the contributions from other data points that are far away, and we derive the GPR model stemming from the application of such localization operation. Through a set of experiments, we demonstrate the competitive performance of the proposed approach compared to full GPR, other localized models, and deep Gaussian processes. Crucially, these performances are obtained with considerable speedups compared to standard global GPR due to the sparsification effect of the Gram matrix induced by the localization operation.

KW - Gaussian processes

KW - kernel smoothing

KW - local regression

UR - http://www.scopus.com/inward/record.url?scp=85143314779&partnerID=8YFLogxK

U2 - 10.1016/j.procs.2022.09.330

DO - 10.1016/j.procs.2022.09.330

M3 - Paper

AN - SCOPUS:85143314779

SP - 2717

EP - 2726

T2 - 26th International Conference on Knowledge-Based and Intelligent Information and Engineering Systems, KES 2022

Y2 - 7 September 2022 through 9 September 2022

ER -

Locally Smoothed Gaussian Process Regression

Abstract

Conference

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this