Presentation | 2016-07-06 Classification analysis of high-dimensional data based on L0-norm optimization. Noriki Ito, Masashi Sato, Yoshiyuki Kabashima, Yoichi Miyawaki, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Advances in sensing devices allow us to measure high-dimensional data easily, but the sample size is often limited because of various reasons such as costs and duration to perform experiments. In such circumstances, feature selection plays a vital role to establish reliable models to describe characteristics of the high-dimensional data. For this purpose, we study iterative algorithms for L0-norm optimization that controls a number of features to be selected. The algorithms have been actively developed for compressed sensing, but not for classification problems explicitly. In this paper, we formulated a classification model with L0-norm regularization based on iterative hard thresholding (IHT) algorithm, quantified its performance in terms of accuracy in classification and feature selection, and compared the performance with that of representative models of a non-sparse classifier (support vector machine) and a sparse classifier (sparse logistic regression). Results showed that the IHT-based classifier outperformed the non-sparse classifier in terms of classification accuracy and did a sparse classifier in terms of feature selection accuracy for certain noise conditions. These results suggest that the proposed model serves an effective means to extract important features embedded in the high-dimensional data. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | feature selection / L0-norm optimization / iterative hard thresholding / high-dimensional data / sparse modeling |
Paper # | NC2016-14 |
Date of Issue | 2016-06-28 (NC) |
Conference Information | |
Committee | NC / IPSJ-BIO / IBISML / IPSJ-MPS |
---|---|
Conference Date | 2016/7/4(3days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Okinawa Institute of Science and Technology |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | Machine Learning Approach to Biodata Mining, and General |
Chair | Shigeo Sato(Tohoku Univ.) / / Kenji Fukumizu(ISM) |
Vice Chair | Masafumi Hagiwara(Keio Univ.) / / Masashi Sugiyama(Univ. of Tokyo) / Hisashi Kashima(Kyoto Univ.) |
Secretary | Masafumi Hagiwara(Kyoto Sangyo Univ.) / (Tokyo Inst. of Tech.) / Masashi Sugiyama / Hisashi Kashima(Univ. of Tokyo) / (Nagoya Inst. of Tech.) |
Assistant | Hisanao Akima(Tohoku Univ.) / Yoshihisa Shinozawa(Keio Univ.) / / Toshihiro Kamishima(AIST) / Tomoharu Iwata(NTT) |
Paper Information | |
Registration To | Technical Committee on Neurocomputing / Special Interest Group on Bioinformatics and Genomics / Technical Committee on Infomation-Based Induction Sciences and Machine Learning / Special Interest Group on Mathematical Modeling and Problem Solving |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Classification analysis of high-dimensional data based on L0-norm optimization. |
Sub Title (in English) | |
Keyword(1) | feature selection |
Keyword(2) | L0-norm optimization |
Keyword(3) | iterative hard thresholding |
Keyword(4) | high-dimensional data |
Keyword(5) | sparse modeling |
1st Author's Name | Noriki Ito |
1st Author's Affiliation | The University of Electro-Communications(UEC Tokyo) |
2nd Author's Name | Masashi Sato |
2nd Author's Affiliation | The University of Electro-Communications(UEC Tokyo) |
3rd Author's Name | Yoshiyuki Kabashima |
3rd Author's Affiliation | Tokyo Institute of Technology(Tokyo Tech) |
4th Author's Name | Yoichi Miyawaki |
4th Author's Affiliation | The University of Electro-Communications(UEC Tokyo) |
Date | 2016-07-06 |
Paper # | NC2016-14 |
Volume (vol) | vol.116 |
Number (no) | NC-120 |
Page | pp.pp.223-228(NC), |
#Pages | 6 |
Date of Issue | 2016-06-28 (NC) |