Paper Abstract and Keywords |
Presentation |
2015-12-21 09:30
Evaluation of Automatic Prototype-Model Size Optimization in Large Geometric Margin Minimum Classification Error Training Masahiro Ogino (Doshisha Univ.), Hideyuki Watanabe (NICT), Shigeru Katagiri, Miho Osaki (Doshisha Univ.), Xugang Lu, Hisashi Kawai (NICT) PRMU2015-100 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
To develop a method for nding an appropriate class model size, which leads to accurate classication over unseen pattern samples, we investigate a method of automatically optimizing the number of prototypes for multi-prototype classiers trained with the Large Geometric Margin Minimum Classication Error training. In
the paper, we propose a new, training-sample-clustering-based procedure of prototype setting, and experimentally demonstrate the utility of the automatic optimization method using the new procedure. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Large geometric margin minimum classication error training / Automatic class model size optimiza- tion / / / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 115, no. 388, PRMU2015-100, pp. 1-6, Dec. 2015. |
Paper # |
PRMU2015-100 |
Date of Issue |
2015-12-14 (PRMU) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
Download PDF |
PRMU2015-100 |
|