Information: Join today and make your research activities more affordable! Technical workshop participation fees and annual registration fees are available at member rates.
Notice: [Important] Announcement of Changes to Registration Fee Payment and Manuscript Upload Procedures for IEICE Technical Meetings
IEICE Technical Committee Submission System
Conference Paper's Information		 Online Proceedings
[Sign in]
Tech. Rep. Archives
 Go Top Page Go Previous   [Japanese] / [English] 

Paper Abstract and Keywords
Presentation 2014-11-18 15:00
[Poster Presentation] Asymptotic Analysis of Variational Bayesian Latent Dirichlet Allocation
Shinichi Nakajima (TU Berlin), Issei Sato, Masashi Sugiyama (Univ. of Tokyo), Kazuho Watanabe (Toyohashi Univ. of Tech.), Hiroko Kobayashi (Nikon) IBISML2014-64
Abstract (in Japanese) (See Japanese page) 
(in English) Latent Dirichlet allocation (LDA) is a popular generative model
of various objects such as texts and images,
where an object is expressed as a mixture of latent topics.
In this paper, we theoretically investigate
variational Bayesian (VB) learning in LDA.
More specifically, we analytically derive
the leading term of the VB free energy under an asymptotic setup,
and show that there exist transition thresholds in Dirichlet hyperparameters
around which the sparsity-inducing behavior drastically changes.
Then we further theoretically reveal the notable phenomenon that
VB tends to induce weaker sparsity than MAP in the LDA model,
which is opposed to other models.
We experimentally demonstrate the practical validity of our asymptotic theory
on real-world Last.FM music data.
Keyword (in Japanese) (See Japanese page) 
(in English) latent Dirichlet allocation / variational Bayes / free energy / partially Bayes / sparsity / / /  
Reference Info. IEICE Tech. Rep., vol. 114, no. 306, IBISML2014-64, pp. 219-226, Nov. 2014.
Paper # IBISML2014-64 
Date of Issue 2014-11-10 (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 IBISML2014-64

Conference Information
Committee IBISML  
Conference Date 2014-11-17 - 2014-11-19 
Place (in Japanese) (See Japanese page) 
Place (in English) Nagoya Univ. 
Topics (in Japanese) (See Japanese page) 
Topics (in English)  
Paper Information
Registration To IBISML 
Conference Code 2014-11-IBISML 
Language English (Japanese title is available) 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Asymptotic Analysis of Variational Bayesian Latent Dirichlet Allocation 
Sub Title (in English)  
Keyword(1) latent Dirichlet allocation  
Keyword(2) variational Bayes  
Keyword(3) free energy  
Keyword(4) partially Bayes  
Keyword(5) sparsity  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Shinichi Nakajima  
1st Author's Affiliation Technical University of Berlin (TU Berlin)
2nd Author's Name Issei Sato  
2nd Author's Affiliation University of Tokyo (Univ. of Tokyo)
3rd Author's Name Masashi Sugiyama  
3rd Author's Affiliation University of Tokyo (Univ. of Tokyo)
4th Author's Name Kazuho Watanabe  
4th Author's Affiliation Toyohashi University of Technology (Toyohashi Univ. of Tech.)
5th Author's Name Hiroko Kobayashi  
5th Author's Affiliation Nikon Corporation (Nikon)
6th Author's Name  
6th Author's Affiliation ()
7th Author's Name  
7th Author's Affiliation ()
8th Author's Name  
8th Author's Affiliation ()
9th Author's Name  
9th Author's Affiliation ()
10th Author's Name  
10th Author's Affiliation ()
11th Author's Name  
11th Author's Affiliation ()
12th Author's Name  
12th Author's Affiliation ()
13th Author's Name  
13th Author's Affiliation ()
14th Author's Name  
14th Author's Affiliation ()
15th Author's Name  
15th Author's Affiliation ()
16th Author's Name  
16th Author's Affiliation ()
17th Author's Name  
17th Author's Affiliation ()
18th Author's Name  
18th Author's Affiliation ()
19th Author's Name  
19th Author's Affiliation ()
20th Author's Name  
20th Author's Affiliation ()
21st Author's Name  
21st Author's Affiliation ()
22nd Author's Name  
22nd Author's Affiliation ()
23rd Author's Name  
23rd Author's Affiliation ()
24th Author's Name  
24th Author's Affiliation ()
25th Author's Name  
25th Author's Affiliation ()
26th Author's Name / /
26th Author's Affiliation ()
()
27th Author's Name / /
27th Author's Affiliation ()
()
28th Author's Name / /
28th Author's Affiliation ()
()
29th Author's Name / /
29th Author's Affiliation ()
()
30th Author's Name / /
30th Author's Affiliation ()
()
31st Author's Name / /
31st Author's Affiliation ()
()
32nd Author's Name / /
32nd Author's Affiliation ()
()
33rd Author's Name / /
33rd Author's Affiliation ()
()
34th Author's Name / /
34th Author's Affiliation ()
()
35th Author's Name / /
35th Author's Affiliation ()
()
36th Author's Name / /
36th Author's Affiliation ()
()
Speaker Author-1 
Date Time 2014-11-18 15:00:00 
Presentation Time 180 minutes 
Registration for IBISML 
Paper # IBISML2014-64 
Volume (vol) vol.114 
Number (no) no.306 
Page pp.219-226 
#Pages 8 
Date of Issue 2014-11-10 (IBISML) 

[Return to Top Page]

[Return to IEICE Web Page]
The Institute of Electronics, Information and Communication Engineers (IEICE), Japan