Multiple kernel learning by conditional entropy minimization

Hideitsu Hino; Nima Reyhani; Noboru Murata

doi:10.1109/ICMLA.2010.40

Multiple kernel learning by conditional entropy minimization

Hideitsu Hino^*
, Nima Reyhani
, Noboru Murata

^*Corresponding author for this work

Waseda University

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

7 Citations (Scopus)

Abstract

Kernel methods have been successfully used in many practical machine learning problems. Choosing a suitable kernel is left to the practitioner. A common way to an automatic selection of optimal kernels is to learn a linear combination of element kernels. In this paper, a novel framework of multiple kernel learning is proposed based on conditional entropy minimization criterion. For the proposed framework, three multiple kernel learning algorithms are derived. The algorithms are experimentally shown to be comparable to or outperform kernel Fisher discriminant analysis and other multiple kernel learning algorithms on benchmark data sets.

Original language	English
Title of host publication	Proceedings - 9th International Conference on Machine Learning and Applications, ICMLA 2010
Pages	223-228
Number of pages	6
DOIs	https://doi.org/10.1109/ICMLA.2010.40
Publication status	Published - 1 Dec 2010
MoE publication type	A4 Conference publication
Event	IEEE International Conference on Machine Learning and Applications - Washington, United States Duration: 12 Dec 2010 → 14 Dec 2010 Conference number: 9

Conference

Conference	IEEE International Conference on Machine Learning and Applications
Abbreviated title	ICMLA
Country/Territory	United States
City	Washington
Period	12/12/2010 → 14/12/2010

Keywords

Discriminant analysis
Entropy
Kernel methods
Multiple Kernel Learning

Access to Document

10.1109/ICMLA.2010.40

Cite this

@inproceedings{006578ef485143b6b9bbdb17cc8168ba,

title = "Multiple kernel learning by conditional entropy minimization",

abstract = "Kernel methods have been successfully used in many practical machine learning problems. Choosing a suitable kernel is left to the practitioner. A common way to an automatic selection of optimal kernels is to learn a linear combination of element kernels. In this paper, a novel framework of multiple kernel learning is proposed based on conditional entropy minimization criterion. For the proposed framework, three multiple kernel learning algorithms are derived. The algorithms are experimentally shown to be comparable to or outperform kernel Fisher discriminant analysis and other multiple kernel learning algorithms on benchmark data sets.",

keywords = "Discriminant analysis, Entropy, Kernel methods, Multiple Kernel Learning",

author = "Hideitsu Hino and Nima Reyhani and Noboru Murata",

year = "2010",

month = dec,

day = "1",

doi = "10.1109/ICMLA.2010.40",

language = "English",

isbn = "9780769543000",

pages = "223--228",

booktitle = "Proceedings - 9th International Conference on Machine Learning and Applications, ICMLA 2010",

note = "IEEE International Conference on Machine Learning and Applications, ICMLA ; Conference date: 12-12-2010 Through 14-12-2010",

}

Hino, H, Reyhani, N & Murata, N 2010, Multiple kernel learning by conditional entropy minimization. in Proceedings - 9th International Conference on Machine Learning and Applications, ICMLA 2010., 5708837, pp. 223-228, IEEE International Conference on Machine Learning and Applications, Washington, District of Columbia, United States, 12/12/2010. https://doi.org/10.1109/ICMLA.2010.40

TY - GEN

T1 - Multiple kernel learning by conditional entropy minimization

AU - Hino, Hideitsu

AU - Reyhani, Nima

AU - Murata, Noboru

N1 - Conference code: 9

PY - 2010/12/1

Y1 - 2010/12/1

N2 - Kernel methods have been successfully used in many practical machine learning problems. Choosing a suitable kernel is left to the practitioner. A common way to an automatic selection of optimal kernels is to learn a linear combination of element kernels. In this paper, a novel framework of multiple kernel learning is proposed based on conditional entropy minimization criterion. For the proposed framework, three multiple kernel learning algorithms are derived. The algorithms are experimentally shown to be comparable to or outperform kernel Fisher discriminant analysis and other multiple kernel learning algorithms on benchmark data sets.

AB - Kernel methods have been successfully used in many practical machine learning problems. Choosing a suitable kernel is left to the practitioner. A common way to an automatic selection of optimal kernels is to learn a linear combination of element kernels. In this paper, a novel framework of multiple kernel learning is proposed based on conditional entropy minimization criterion. For the proposed framework, three multiple kernel learning algorithms are derived. The algorithms are experimentally shown to be comparable to or outperform kernel Fisher discriminant analysis and other multiple kernel learning algorithms on benchmark data sets.

KW - Discriminant analysis

KW - Entropy

KW - Kernel methods

KW - Multiple Kernel Learning

UR - https://www.scopus.com/pages/publications/79952420541

U2 - 10.1109/ICMLA.2010.40

DO - 10.1109/ICMLA.2010.40

M3 - Conference article in proceedings

AN - SCOPUS:79952420541

SN - 9780769543000

SP - 223

EP - 228

BT - Proceedings - 9th International Conference on Machine Learning and Applications, ICMLA 2010

T2 - IEEE International Conference on Machine Learning and Applications

Y2 - 12 December 2010 through 14 December 2010

ER -

Multiple kernel learning by conditional entropy minimization

Abstract

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this