Massively-parallel best subset selection for ordinary least-squares regression

Fabian Gieseke; Kai Lars Polsterer; Ashish Mahabal; Christian Igel; Tom Heskes

doi:10.1109/SSCI.2017.8285225

Massively-parallel best subset selection for ordinary least-squares regression

Fabian Gieseke, Kai Lars Polsterer, Ashish Mahabal, Christian Igel, Tom Heskes

Datalogisk Institut

2 Citationer (Scopus)

Abstract

Selecting an optimal subset of k out of d features for linear regression models given n training instances is often considered intractable for feature spaces with hundreds or thousands of dimensions. We propose an efficient massively-parallel implementation for selecting such optimal feature subsets in a brute-force fashion for small k. By exploiting the enormous compute power provided by modern parallel devices such as graphics processing units, it can deal with thousands of input dimensions even using standard commodity hardware only. We evaluate the practical runtime using artificial datasets and sketch the applicability of our framework in the context of astronomy.

Originalsprog	Engelsk
Titel	2017 IEEE Symposium Series on Computational Intelligence (SSCI) Proceedings
Antal sider	8
Forlag	IEEE
Publikationsdato	1 jul. 2017
Sider	1-8
ISBN (Elektronisk)	978-1-5386-2726-6
DOI	https://doi.org/10.1109/SSCI.2017.8285225
Status	Udgivet - 1 jul. 2017
Begivenhed	2017 IEEE Symposium Series on Computational Intelligence (SSCI) - Honolulu, USA Varighed: 27 nov. 2017 → 1 dec. 2017

Konference

Konference	2017 IEEE Symposium Series on Computational Intelligence (SSCI)
Land/Område	USA
By	Honolulu
Periode	27/11/2017 → 01/12/2017

Adgang til dokumentet

10.1109/SSCI.2017.8285225

https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8285225

Citationsformater

Gieseke, F, Polsterer, KL, Mahabal, A, Igel, C & Heskes, T 2017, Massively-parallel best subset selection for ordinary least-squares regression. i 2017 IEEE Symposium Series on Computational Intelligence (SSCI) Proceedings. IEEE, s. 1-8, 2017 IEEE Symposium Series on Computational Intelligence (SSCI), Honolulu, USA, 27/11/2017. https://doi.org/10.1109/SSCI.2017.8285225

@inproceedings{86ffbdbd2e8d4764bee8902f9922f3fc,

title = "Massively-parallel best subset selection for ordinary least-squares regression",

abstract = "Selecting an optimal subset of k out of d features for linear regression models given n training instances is often considered intractable for feature spaces with hundreds or thousands of dimensions. We propose an efficient massively-parallel implementation for selecting such optimal feature subsets in a brute-force fashion for small k. By exploiting the enormous compute power provided by modern parallel devices such as graphics processing units, it can deal with thousands of input dimensions even using standard commodity hardware only. We evaluate the practical runtime using artificial datasets and sketch the applicability of our framework in the context of astronomy.",

keywords = "graphics processing units, least squares approximations, optimisation, parallel processing, regression analysis, sensitivity analysis, input dimensions, linear regression models, massively-parallel best subset selection, optimal feature subsets, optimal subset, ordinary least-squares regression, subset selection, Computational modeling, Graphics processing units, Instruction sets, Optimization, Runtime, Task analysis, Training",

author = "Fabian Gieseke and Polsterer, {Kai Lars} and Ashish Mahabal and Christian Igel and Tom Heskes",

year = "2017",

month = jul,

day = "1",

doi = "10.1109/SSCI.2017.8285225",

language = "English",

pages = "1--8",

booktitle = "2017 IEEE Symposium Series on Computational Intelligence (SSCI) Proceedings",

publisher = "IEEE",

note = "2017 IEEE Symposium Series on Computational Intelligence (SSCI) ; Conference date: 27-11-2017 Through 01-12-2017",

}

TY - GEN

T1 - Massively-parallel best subset selection for ordinary least-squares regression

AU - Gieseke, Fabian

AU - Polsterer, Kai Lars

AU - Mahabal, Ashish

AU - Igel, Christian

AU - Heskes, Tom

PY - 2017/7/1

Y1 - 2017/7/1

N2 - Selecting an optimal subset of k out of d features for linear regression models given n training instances is often considered intractable for feature spaces with hundreds or thousands of dimensions. We propose an efficient massively-parallel implementation for selecting such optimal feature subsets in a brute-force fashion for small k. By exploiting the enormous compute power provided by modern parallel devices such as graphics processing units, it can deal with thousands of input dimensions even using standard commodity hardware only. We evaluate the practical runtime using artificial datasets and sketch the applicability of our framework in the context of astronomy.

AB - Selecting an optimal subset of k out of d features for linear regression models given n training instances is often considered intractable for feature spaces with hundreds or thousands of dimensions. We propose an efficient massively-parallel implementation for selecting such optimal feature subsets in a brute-force fashion for small k. By exploiting the enormous compute power provided by modern parallel devices such as graphics processing units, it can deal with thousands of input dimensions even using standard commodity hardware only. We evaluate the practical runtime using artificial datasets and sketch the applicability of our framework in the context of astronomy.

KW - graphics processing units

KW - least squares approximations

KW - optimisation

KW - parallel processing

KW - regression analysis

KW - sensitivity analysis

KW - input dimensions

KW - linear regression models

KW - massively-parallel best subset selection

KW - optimal feature subsets

KW - optimal subset

KW - ordinary least-squares regression

KW - subset selection

KW - Computational modeling

KW - Graphics processing units

KW - Instruction sets

KW - Optimization

KW - Runtime

KW - Task analysis

KW - Training

U2 - 10.1109/SSCI.2017.8285225

DO - 10.1109/SSCI.2017.8285225

M3 - Article in proceedings

SP - 1

EP - 8

BT - 2017 IEEE Symposium Series on Computational Intelligence (SSCI) Proceedings

PB - IEEE

T2 - 2017 IEEE Symposium Series on Computational Intelligence (SSCI)

Y2 - 27 November 2017 through 1 December 2017

ER -

Massively-parallel best subset selection for ordinary least-squares regression

Abstract

Konference

Adgang til dokumentet

Fingeraftryk

Citationsformater