Paper Abstract and Keywords |
Presentation |
2011-09-05 10:30
On Evaluation of Stochastic Complexity based on Bayes Code and Its Applications to Model Selection Yoshinari Takeishi, Masanori Kawakita, Jun'ichi Takeuchi (Kyushu Univ./ISIT) PRMU2011-59 IBISML2011-18 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
We evaluate stochastic complexity of Gaussian mixture by Bayes code length, and apply it to the model selection problem.The Gaussian mixture is a popular model in machine learning and statistics. Since it is generally difficult to calculate Bayes code length of Gaussian mixture, we propose an accurate approximation method to calculate it in practical time, which uses Laplace approximation and Monte Carlo method. We compare the performance of our method with other methods. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Gaussian Mixture / Stochastic Complexity / Bayes Code / Model Selection / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 111, no. 194, IBISML2011-18, pp. 9-14, Sept. 2011. |
Paper # |
IBISML2011-18 |
Date of Issue |
2011-08-29 (PRMU, IBISML) |
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 |
PRMU2011-59 IBISML2011-18 |
|