Paper Abstract and Keywords |
Presentation |
2009-08-31 10:00
Geometric Margin for a General Class of Discriminant Functions and Its Control for Minimum Error Classification Hideyuki Watanabe (NICT), Shigeru Katagiri, Kouta Yamada (Doshisha Univ.), Erik McDermott, Atsushi Nakamura, Shinji Watanabe (NTT Corp.), Miho Ohsaki (Doshisha Univ.) PRMU2009-60 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
The recent dramatic growth of computation power has resulted in increased interests in discriminative training (DT) methods such as Minimum Classification Error (MCE) training and Support Vector Machine (SVM) training for pattern recognition. MCE is a general DT framework that can be used to achieve minimum error classification of various types of patterns. SVM can be used to improve the classification robustness for mainly classifying fixed-dimensional patterns; this is done by maximizing the geometric margin for a class of linear discriminant functions. To realize high robustness in a wide range of classification tasks, we derive the geometric margin for a general class of discriminant functions and develop a new MCE training method, in which the geometric margin is set to a large value. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Minimum Classification Error training / MCE / Support Vector Machine / margin / geometric margin / / / |
Reference Info. |
IEICE Tech. Rep., vol. 109, no. 182, PRMU2009-60, pp. 1-6, Aug. 2009. |
Paper # |
PRMU2009-60 |
Date of Issue |
2009-08-24 (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 |
PRMU2009-60 |
|