Presentation | 2007-03-14 Split Parameter Optimization for Multiclass Support Vector Machines Keita TSUBOTA, Mauricio KUGLER, NUGROHO Anto SATRIYO, Susumu KUROYANAGI, Akira IWATA, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | A support vector machine (SVM) using Gaussian kernel function presents two parameters (C and σ) that need to be tuned in order to obtain good generalization. Several automatic parameters tuning techniques based on the classifierc's recognition rate have been proposed for a single binary support vector machine. However, the naive use of the multiclass SVM recognition rate as a criterion for searching good parameters becomes unpractical when dealing with large-scale classification problems due to the high processing time required. Specific methods for multiclass SVM parameters optimization are not widely studied. This paper proposes a new approach for the optimization of multiclass SVM parameters. The new method splits the optimization procedure, searching for independent parameters for each binary classifier. The experimental results show that, while keeping the same accuracy performance, the proposed algorithm significantly decreases the optimization procedure's processing time and reduces the total final amount of support vectors. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Support vector machine / parameters tuning / malitclass classification / large-scale data |
Paper # | NC2006-138 |
Date of Issue |
Conference Information | |
Committee | NC |
---|---|
Conference Date | 2007/3/7(1days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | |
Chair | |
Vice Chair | |
Secretary | |
Assistant |
Paper Information | |
Registration To | Neurocomputing (NC) |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Split Parameter Optimization for Multiclass Support Vector Machines |
Sub Title (in English) | |
Keyword(1) | Support vector machine |
Keyword(2) | parameters tuning |
Keyword(3) | malitclass classification |
Keyword(4) | large-scale data |
1st Author's Name | Keita TSUBOTA |
1st Author's Affiliation | Department of Computer Science & Engineering, Nagoya Institute of Technology() |
2nd Author's Name | Mauricio KUGLER |
2nd Author's Affiliation | Department of Computer Science & Engineering, Nagoya Institute of Technology |
3rd Author's Name | NUGROHO Anto SATRIYO |
3rd Author's Affiliation | School of Life Science & Technology, Chukyo University |
4th Author's Name | Susumu KUROYANAGI |
4th Author's Affiliation | Department of Computer Science & Engineering, Nagoya Institute of Technology |
5th Author's Name | Akira IWATA |
5th Author's Affiliation | Department of Computer Science & Engineering, Nagoya Institute of Technology |
Date | 2007-03-14 |
Paper # | NC2006-138 |
Volume (vol) | vol.106 |
Number (no) | 588 |
Page | pp.pp.- |
#Pages | 6 |
Date of Issue |