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 2022-01-23 11:45
Adversarial Training with Knowledge Distillation considering Intermediate Feature Representation in CNNs
Hikaru Higuchi (The Univ. of Electro-Communications), Satoshi Suzuki (former NTT), Hayaru Shouno (The Univ. of Electro-Communications) NC2021-44
Abstract (in Japanese) (See Japanese page) 
(in English) Adversarial examples are one of the vulnerability attacks to the convolution neural network (CNN). The adversarialexamples are made by adding adversarial perturbations, which are maliciously designed to deceive the target DNN and aregenerally human-imperceptible, to input images. Adversarial training is a method to improve classification accuracy againstadversarial attacks. In the adversarial training, the CNN is trained with not clean images (not including adversarial pertur-bations) but adversarial examples. However, conventional adversarial training decreases the classification accuracy on cleanimages than usual training which trains the CNN with clean images only. From our experimental results, the CNNs trained onclean images only can obtain effective feature representations for classifying clean images, while the conventional adversarialtraining cannot. In accordance with this perspective, we propose a new adversarial training method based on knowledgedistillation using clean-CNN that trained with clean images only as a teacher model. This method transfers the knowledge fromthe clean-CNN and makes feature representations effective for classifying clean images in adversarial training. Our methodoutperforms the conventional adversarial training for both clean images and adversarial examples.
Keyword (in Japanese) (See Japanese page) 
(in English) Convolutional Neural Network / Adversarial Training / Knowledge Distillation / Manifold Hypothesis / / / /  
Reference Info. IEICE Tech. Rep., vol. 121, no. 338, NC2021-44, pp. 59-64, Jan. 2022.
Paper # NC2021-44 
Date of Issue 2022-01-14 (NC) 
ISSN 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 NC2021-44

Conference Information
Committee NLP MICT MBE NC  
Conference Date 2022-01-21 - 2022-01-23 
Place (in Japanese) (See Japanese page) 
Place (in English) Online 
Topics (in Japanese) (See Japanese page) 
Topics (in English)  
Paper Information
Registration To NC 
Conference Code 2022-01-NLP-MICT-MBE-NC 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Adversarial Training with Knowledge Distillation considering Intermediate Feature Representation in CNNs 
Sub Title (in English)  
Keyword(1) Convolutional Neural Network  
Keyword(2) Adversarial Training  
Keyword(3) Knowledge Distillation  
Keyword(4) Manifold Hypothesis  
Keyword(5)  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Hikaru Higuchi  
1st Author's Affiliation The University of Electro-Communications (The Univ. of Electro-Communications)
2nd Author's Name Satoshi Suzuki  
2nd Author's Affiliation NTT Computer and Data Science Laboratories, NTT Corporation (former NTT)
3rd Author's Name Hayaru Shouno  
3rd Author's Affiliation The University of Electro-Communications (The Univ. of Electro-Communications)
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 ()
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 2022-01-23 11:45:00 
Presentation Time 25 minutes 
Registration for NC 
Paper # NC2021-44 
Volume (vol) vol.121 
Number (no) no.338 
Page pp.59-64 
#Pages 6 
Date of Issue 2022-01-14 (NC) 

[Return to Top Page]

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