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 2016-11-17 14:00
[Poster Presentation] Analysis of Multimodal Deep Neural Networks -- Towards the elucidation of the modality integration mechanism --
Yoh-ichi Mototake, Takashi Ikegami (unit of Tokyo) IBISML2016-97
Abstract (in Japanese) (See Japanese page) 
(in English) With the rapid development of information technology in recent years,
several machine learning algorithms that integrate some information which have different modalities
such as images, texts, and sounds for generating sentences and any other have been put to practical use.
On the other hand, little is known about the mechanisms by which our modalities are integrated.
The purpose of this study is to explore the mechanisms through the analysis of the internal state
of Multimodal Deep Neural Networks(DNNs) whose learning algorithms are rapidly developed in recent years.
First, we assumed that datasets have structures based on the manifold hypothesis,
and then computed the manifolds' tangent spaces from the mapping function of DNNs.
Finally, we calculated how the geometric structures were converted inside the DNNs.
The result of the analysis suggested that Multimodal DNNs have functions
to map manifold structures of the distribution of datasets to a global coordinate system, and
that is processed after the integration of the modalities.
Keyword (in Japanese) (See Japanese page) 
(in English) Multimodal Deep Learning / Manifold Hypothesis / / / / / /  
Reference Info. IEICE Tech. Rep., vol. 116, no. 300, IBISML2016-97, pp. 369-373, Nov. 2016.
Paper # IBISML2016-97 
Date of Issue 2016-11-09 (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 IBISML2016-97

Conference Information
Committee IBISML  
Conference Date 2016-11-16 - 2016-11-18 
Place (in Japanese) (See Japanese page) 
Place (in English) Kyoto Univ. 
Topics (in Japanese) (See Japanese page) 
Topics (in English) Information-Based Induction Science Workshop (IBIS2016) 
Paper Information
Registration To IBISML 
Conference Code 2016-11-IBISML 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Analysis of Multimodal Deep Neural Networks 
Sub Title (in English) Towards the elucidation of the modality integration mechanism 
Keyword(1) Multimodal Deep Learning  
Keyword(2) Manifold Hypothesis  
Keyword(3)  
Keyword(4)  
Keyword(5)  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Yoh-ichi Mototake  
1st Author's Affiliation University of Tokyo (unit of Tokyo)
2nd Author's Name Takashi Ikegami  
2nd Author's Affiliation University of Tokyo (unit of Tokyo)
3rd Author's Name  
3rd Author's Affiliation ()
4th Author's Name  
4th Author's Affiliation ()
5th Author's Name  
5th Author's Affiliation ()
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 ()
Speaker Author-1 
Date Time 2016-11-17 14:00:00 
Presentation Time 180 minutes 
Registration for IBISML 
Paper # IBISML2016-97 
Volume (vol) vol.116 
Number (no) no.300 
Page pp.369-373 
#Pages
Date of Issue 2016-11-09 (IBISML) 


[Return to Top Page]

[Return to IEICE Web Page]


The Institute of Electronics, Information and Communication Engineers (IEICE), Japan