Topology potential-based parameter selecting for support vector machine

Yi Lin; Shuliang Wang; Long Zhao; Da Kui Wang

doi:10.1007/978-3-319-14717-8_40

Topology potential-based parameter selecting for support vector machine

Yi Lin, Shuliang Wang^*, Long Zhao, Da Kui Wang

^*Corresponding author for this work

School of Computer Science and Technology

Wuhan University

Research output: Contribution to journal › Article › peer-review

2 Citations (Scopus)

Abstract

We present an algorithm for selecting support vector machine’s meta- parameter value which is based on ideas from topology potential of data field. By the optimal spatial distribution of topological potential corresponding to minimum entropy potential, it searches so smart that the optimal parameters can be found effectively and efficiently. The experimental results show that it can get almost the same effectiveness with the exhaustive grid search under an order of magnitude lower computational cost. It also can be used to automatically identify kernels and other parameter selection problem.

Original language	English
Pages (from-to)	513-522
Number of pages	10
Journal	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	8933
DOIs	https://doi.org/10.1007/978-3-319-14717-8_40
Publication status	Published - 2014

Keywords

Data field
Parameter selection
Potential entropy
Support vector machine
Topology potential

Access to Document

10.1007/978-3-319-14717-8_40

Cite this

@article{7781d8e0bdbb495d8e0234cea624c2c1,

title = "Topology potential-based parameter selecting for support vector machine",

abstract = "We present an algorithm for selecting support vector machine{\textquoteright}s meta- parameter value which is based on ideas from topology potential of data field. By the optimal spatial distribution of topological potential corresponding to minimum entropy potential, it searches so smart that the optimal parameters can be found effectively and efficiently. The experimental results show that it can get almost the same effectiveness with the exhaustive grid search under an order of magnitude lower computational cost. It also can be used to automatically identify kernels and other parameter selection problem.",

keywords = "Data field, Parameter selection, Potential entropy, Support vector machine, Topology potential",

author = "Yi Lin and Shuliang Wang and Long Zhao and Wang, {Da Kui}",

note = "Publisher Copyright: {\textcopyright} Springer International Publishing Switzerland 2014.",

year = "2014",

doi = "10.1007/978-3-319-14717-8_40",

language = "English",

volume = "8933",

pages = "513--522",

journal = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

issn = "0302-9743",

publisher = "Springer Science and Business Media Deutschland GmbH",

}

TY - JOUR

T1 - Topology potential-based parameter selecting for support vector machine

AU - Lin, Yi

AU - Wang, Shuliang

AU - Zhao, Long

AU - Wang, Da Kui

N1 - Publisher Copyright: © Springer International Publishing Switzerland 2014.

PY - 2014

Y1 - 2014

N2 - We present an algorithm for selecting support vector machine’s meta- parameter value which is based on ideas from topology potential of data field. By the optimal spatial distribution of topological potential corresponding to minimum entropy potential, it searches so smart that the optimal parameters can be found effectively and efficiently. The experimental results show that it can get almost the same effectiveness with the exhaustive grid search under an order of magnitude lower computational cost. It also can be used to automatically identify kernels and other parameter selection problem.

AB - We present an algorithm for selecting support vector machine’s meta- parameter value which is based on ideas from topology potential of data field. By the optimal spatial distribution of topological potential corresponding to minimum entropy potential, it searches so smart that the optimal parameters can be found effectively and efficiently. The experimental results show that it can get almost the same effectiveness with the exhaustive grid search under an order of magnitude lower computational cost. It also can be used to automatically identify kernels and other parameter selection problem.

KW - Data field

KW - Parameter selection

KW - Potential entropy

KW - Support vector machine

KW - Topology potential

UR - http://www.scopus.com/inward/record.url?scp=84918823593&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-14717-8_40

DO - 10.1007/978-3-319-14717-8_40

M3 - Article

AN - SCOPUS:84918823593

SN - 0302-9743

VL - 8933

SP - 513

EP - 522

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

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

ER -

Topology potential-based parameter selecting for support vector machine

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this