Localized multiple kernel learning

Mehmet Gonen, Ethem Alpaydin

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

189 Citations (Scopus)

Abstract

Recently, instead of selecting a single kernel, multiple kernel learning (MKL) has been proposed which uses a convex combination of kernels, where the weight of each kernel is optimized during training. However, MKL assigns the same weight to a kernel over the whole input space. In this paper, we develop a localized multiple kernel learning (LMKL) algorithm using a gating model for selecting the appropriate kernel function locally. The localizing gating model and the kernel-based classifier are coupled and their optimization is done in a joint mariner. Empirical results on ten benchmark and two bioinformatics data sets validate the applicability of our approach. LMKL achieves statistically similar accuracy results compared with MKL by storing fewer support vectors. LMKL can also combine multiple copies of the same kernel function localized in different parts. For example, LMKL with multiple linear kernels gives better accuracy results than using a single linear kernel on bioinformatics data sets.

Original languageEnglish (US)
Title of host publicationProceedings of the 25th International Conference on Machine Learning
Pages352-359
Number of pages8
StatePublished - 2008
Externally publishedYes
Event25th International Conference on Machine Learning - Helsinki, Finland
Duration: Jul 5 2008Jul 9 2008

Other

Other25th International Conference on Machine Learning
CountryFinland
CityHelsinki
Period7/5/087/9/08

Fingerprint

Bioinformatics
Learning algorithms
Classifiers

ASJC Scopus subject areas

  • Artificial Intelligence
  • Human-Computer Interaction
  • Software

Cite this

Gonen, M., & Alpaydin, E. (2008). Localized multiple kernel learning. In Proceedings of the 25th International Conference on Machine Learning (pp. 352-359)

Localized multiple kernel learning. / Gonen, Mehmet; Alpaydin, Ethem.

Proceedings of the 25th International Conference on Machine Learning. 2008. p. 352-359.

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

Gonen, M & Alpaydin, E 2008, Localized multiple kernel learning. in Proceedings of the 25th International Conference on Machine Learning. pp. 352-359, 25th International Conference on Machine Learning, Helsinki, Finland, 7/5/08.
Gonen M, Alpaydin E. Localized multiple kernel learning. In Proceedings of the 25th International Conference on Machine Learning. 2008. p. 352-359
Gonen, Mehmet ; Alpaydin, Ethem. / Localized multiple kernel learning. Proceedings of the 25th International Conference on Machine Learning. 2008. pp. 352-359
@inproceedings{7bb50fc1c1264895afedb67144365bc4,
title = "Localized multiple kernel learning",
abstract = "Recently, instead of selecting a single kernel, multiple kernel learning (MKL) has been proposed which uses a convex combination of kernels, where the weight of each kernel is optimized during training. However, MKL assigns the same weight to a kernel over the whole input space. In this paper, we develop a localized multiple kernel learning (LMKL) algorithm using a gating model for selecting the appropriate kernel function locally. The localizing gating model and the kernel-based classifier are coupled and their optimization is done in a joint mariner. Empirical results on ten benchmark and two bioinformatics data sets validate the applicability of our approach. LMKL achieves statistically similar accuracy results compared with MKL by storing fewer support vectors. LMKL can also combine multiple copies of the same kernel function localized in different parts. For example, LMKL with multiple linear kernels gives better accuracy results than using a single linear kernel on bioinformatics data sets.",
author = "Mehmet Gonen and Ethem Alpaydin",
year = "2008",
language = "English (US)",
isbn = "9781605582054",
pages = "352--359",
booktitle = "Proceedings of the 25th International Conference on Machine Learning",

}

TY - GEN

T1 - Localized multiple kernel learning

AU - Gonen, Mehmet

AU - Alpaydin, Ethem

PY - 2008

Y1 - 2008

N2 - Recently, instead of selecting a single kernel, multiple kernel learning (MKL) has been proposed which uses a convex combination of kernels, where the weight of each kernel is optimized during training. However, MKL assigns the same weight to a kernel over the whole input space. In this paper, we develop a localized multiple kernel learning (LMKL) algorithm using a gating model for selecting the appropriate kernel function locally. The localizing gating model and the kernel-based classifier are coupled and their optimization is done in a joint mariner. Empirical results on ten benchmark and two bioinformatics data sets validate the applicability of our approach. LMKL achieves statistically similar accuracy results compared with MKL by storing fewer support vectors. LMKL can also combine multiple copies of the same kernel function localized in different parts. For example, LMKL with multiple linear kernels gives better accuracy results than using a single linear kernel on bioinformatics data sets.

AB - Recently, instead of selecting a single kernel, multiple kernel learning (MKL) has been proposed which uses a convex combination of kernels, where the weight of each kernel is optimized during training. However, MKL assigns the same weight to a kernel over the whole input space. In this paper, we develop a localized multiple kernel learning (LMKL) algorithm using a gating model for selecting the appropriate kernel function locally. The localizing gating model and the kernel-based classifier are coupled and their optimization is done in a joint mariner. Empirical results on ten benchmark and two bioinformatics data sets validate the applicability of our approach. LMKL achieves statistically similar accuracy results compared with MKL by storing fewer support vectors. LMKL can also combine multiple copies of the same kernel function localized in different parts. For example, LMKL with multiple linear kernels gives better accuracy results than using a single linear kernel on bioinformatics data sets.

UR - http://www.scopus.com/inward/record.url?scp=56449124689&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=56449124689&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:56449124689

SN - 9781605582054

SP - 352

EP - 359

BT - Proceedings of the 25th International Conference on Machine Learning

ER -