Multitask learning using regularized multiple kernel learning

Mehmet Gonen, Melih Kandemir, Samuel Kaski

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

3 Citations (Scopus)

Abstract

Empirical success of kernel-based learning algorithms is very much dependent on the kernel function used. Instead of using a single fixed kernel function, multiple kernel learning (MKL) algorithms learn a combination of different kernel functions in order to obtain a similarity measure that better matches the underlying problem. We study multitask learning (MKL) problems and formulate a novel MTL algorithm that trains coupled but nonidentical MKL models across the tasks. The proposed algorithm is especially useful for tasks that have different input and/or output space characteristics and is computationally very efficient. Empirical results on three data sets validate the generalization performance and the efficiency of our approach.

Original languageEnglish (US)
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages500-509
Number of pages10
Volume7063 LNCS
EditionPART 2
DOIs
StatePublished - 2011
Externally publishedYes
Event18th International Conference on Neural Information Processing, ICONIP 2011 - Shanghai, China
Duration: Nov 13 2011Nov 17 2011

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 2
Volume7063 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other18th International Conference on Neural Information Processing, ICONIP 2011
CountryChina
CityShanghai
Period11/13/1111/17/11

Fingerprint

Multi-task Learning
Kernel Function
kernel
Learning algorithms
Learning Algorithm
Similarity Measure
Dependent
Output
Learning
Model

Keywords

  • kernel machines
  • multilabel learning
  • multiple kernel learning
  • multitask learning
  • support vector machines

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Gonen, M., Kandemir, M., & Kaski, S. (2011). Multitask learning using regularized multiple kernel learning. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (PART 2 ed., Vol. 7063 LNCS, pp. 500-509). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7063 LNCS, No. PART 2). https://doi.org/10.1007/978-3-642-24958-7_58

Multitask learning using regularized multiple kernel learning. / Gonen, Mehmet; Kandemir, Melih; Kaski, Samuel.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 7063 LNCS PART 2. ed. 2011. p. 500-509 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7063 LNCS, No. PART 2).

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

Gonen, M, Kandemir, M & Kaski, S 2011, Multitask learning using regularized multiple kernel learning. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 2 edn, vol. 7063 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 2, vol. 7063 LNCS, pp. 500-509, 18th International Conference on Neural Information Processing, ICONIP 2011, Shanghai, China, 11/13/11. https://doi.org/10.1007/978-3-642-24958-7_58
Gonen M, Kandemir M, Kaski S. Multitask learning using regularized multiple kernel learning. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 2 ed. Vol. 7063 LNCS. 2011. p. 500-509. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 2). https://doi.org/10.1007/978-3-642-24958-7_58
Gonen, Mehmet ; Kandemir, Melih ; Kaski, Samuel. / Multitask learning using regularized multiple kernel learning. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 7063 LNCS PART 2. ed. 2011. pp. 500-509 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 2).
@inproceedings{e450e89f36c242de81193617bbf17411,
title = "Multitask learning using regularized multiple kernel learning",
abstract = "Empirical success of kernel-based learning algorithms is very much dependent on the kernel function used. Instead of using a single fixed kernel function, multiple kernel learning (MKL) algorithms learn a combination of different kernel functions in order to obtain a similarity measure that better matches the underlying problem. We study multitask learning (MKL) problems and formulate a novel MTL algorithm that trains coupled but nonidentical MKL models across the tasks. The proposed algorithm is especially useful for tasks that have different input and/or output space characteristics and is computationally very efficient. Empirical results on three data sets validate the generalization performance and the efficiency of our approach.",
keywords = "kernel machines, multilabel learning, multiple kernel learning, multitask learning, support vector machines",
author = "Mehmet Gonen and Melih Kandemir and Samuel Kaski",
year = "2011",
doi = "10.1007/978-3-642-24958-7_58",
language = "English (US)",
isbn = "9783642249570",
volume = "7063 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
number = "PART 2",
pages = "500--509",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
edition = "PART 2",

}

TY - GEN

T1 - Multitask learning using regularized multiple kernel learning

AU - Gonen, Mehmet

AU - Kandemir, Melih

AU - Kaski, Samuel

PY - 2011

Y1 - 2011

N2 - Empirical success of kernel-based learning algorithms is very much dependent on the kernel function used. Instead of using a single fixed kernel function, multiple kernel learning (MKL) algorithms learn a combination of different kernel functions in order to obtain a similarity measure that better matches the underlying problem. We study multitask learning (MKL) problems and formulate a novel MTL algorithm that trains coupled but nonidentical MKL models across the tasks. The proposed algorithm is especially useful for tasks that have different input and/or output space characteristics and is computationally very efficient. Empirical results on three data sets validate the generalization performance and the efficiency of our approach.

AB - Empirical success of kernel-based learning algorithms is very much dependent on the kernel function used. Instead of using a single fixed kernel function, multiple kernel learning (MKL) algorithms learn a combination of different kernel functions in order to obtain a similarity measure that better matches the underlying problem. We study multitask learning (MKL) problems and formulate a novel MTL algorithm that trains coupled but nonidentical MKL models across the tasks. The proposed algorithm is especially useful for tasks that have different input and/or output space characteristics and is computationally very efficient. Empirical results on three data sets validate the generalization performance and the efficiency of our approach.

KW - kernel machines

KW - multilabel learning

KW - multiple kernel learning

KW - multitask learning

KW - support vector machines

UR - http://www.scopus.com/inward/record.url?scp=81855227233&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=81855227233&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-24958-7_58

DO - 10.1007/978-3-642-24958-7_58

M3 - Conference contribution

AN - SCOPUS:81855227233

SN - 9783642249570

VL - 7063 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 500

EP - 509

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

ER -