Paper Abstract and Keywords |
Presentation |
2016-01-29 16:15
Proposal of novel dropout method and its analysis of dynamic property Daisuke Saitoh, Tasuku Kondo, Kazuyuki Hara (Nihon Univ.) NC2015-67 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
Deep learning that use a large network and includes many units tends to occur the overfitting. Therefore, to avoid the overfitting, several regularization methods have been proposed, and one of them is Dropout. Dropout selects some units of the network at random and it drops them in the learning process. Then the network size becomes smaller, and the overfitting can be avoided. When calculating the output, we sum up the dropped units and that of not dropped units. This seems like the ensemble learning. On the other hand, there is a phenomenon called the symmetry breaking that achieves very small residual error. Then, we treated Dropout as the ensemble learning, and explored the Dropout that achieves the symmetry breaking. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Dropout / regularization / online learning / neural networks / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 115, no. 426, NC2015-67, pp. 55-60, Jan. 2016. |
Paper # |
NC2015-67 |
Date of Issue |
2016-01-21 (NC) |
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 |
NC2015-67 |
Conference Information |
Committee |
NC NLP |
Conference Date |
2016-01-28 - 2016-01-29 |
Place (in Japanese) |
(See Japanese page) |
Place (in English) |
Kyushu Institute of Technology |
Topics (in Japanese) |
(See Japanese page) |
Topics (in English) |
Implementation of Neuro Computing,Analysis and Modeling of Human Science, etc |
Paper Information |
Registration To |
NC |
Conference Code |
2016-01-NC-NLP |
Language |
Japanese |
Title (in Japanese) |
(See Japanese page) |
Sub Title (in Japanese) |
(See Japanese page) |
Title (in English) |
Proposal of novel dropout method and its analysis of dynamic property |
Sub Title (in English) |
|
Keyword(1) |
Dropout |
Keyword(2) |
regularization |
Keyword(3) |
online learning |
Keyword(4) |
neural networks |
Keyword(5) |
|
Keyword(6) |
|
Keyword(7) |
|
Keyword(8) |
|
1st Author's Name |
Daisuke Saitoh |
1st Author's Affiliation |
Nihon University (Nihon Univ.) |
2nd Author's Name |
Tasuku Kondo |
2nd Author's Affiliation |
Nihon University (Nihon Univ.) |
3rd Author's Name |
Kazuyuki Hara |
3rd Author's Affiliation |
Nihon University (Nihon Univ.) |
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 |
2016-01-29 16:15:00 |
Presentation Time |
25 minutes |
Registration for |
NC |
Paper # |
NC2015-67 |
Volume (vol) |
vol.115 |
Number (no) |
no.426 |
Page |
pp.55-60 |
#Pages |
6 |
Date of Issue |
2016-01-21 (NC) |
|