Paper Abstract and Keywords |
Presentation |
2009-09-01 09:00
Implementation and Experimental Evaluation of Ensemble Minimum Classification Error Training Shin'ichi Taniguchi (Doshisha University), Hideyuki Watanabe (NICT), Shigeru Katagiri, Kohta Yamada (Doshisha University), Atsushi Nakamura, Erik McDermott, Shinji Watanabe (NTT), Naho Nishijima, Miho Ohsaki (Doshisha Univ.) PRMU2009-67 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
Recently, we developed a noble Ensemble-based Minimum Classification Error training method (EMCE) by combining the advantages of Boosting, which is a typical ensemble-based classifier design method, and the MCE method that has been widely used in the speech recognition and text processing field. In this paper, we ellaborate the implementation and experimental evaluation of EMCE. Through systematic experiments, we show the superiority of EMCE to its conventional counterpart, Boosting, and we also discuss how EMCE should be further improved. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Minimum classification error training / Discriminative training / Boosting / Ensemble training / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 109, no. 182, PRMU2009-67, pp. 103-108, Aug. 2009. |
Paper # |
PRMU2009-67 |
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 |
PRMU2009-67 |
|