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 2026-03-24 15:20
Improving the Robustness/Accuracy Tradeoff in Robust Information Bottleneck Distillation Through Dual Teachers
Vincent Ryusuke Takahashi, Yoshinari Takeishi, Jun'ichi Takeuchi (Kyushu Univ.), Kave Salamatian (Savoie Mont Blanc Univ.) IBISML2025-44
Abstract (in Japanese) (See Japanese page) 
(in English) Deep neural networks (DNNs) have achieved remarkable success in classical machine learning problems. However, they are known to be vulnerable to adversarial attacks. Countermeasures proposed in the literature, notably Information Bottleneck Distillation (IBD) introduced by Kuang et al., degrades the classification accuracy on clean inputs while improving the robustness to adversarial inputs. In this work, we extend the IBD framework by introducing in the distillation an additional “clean teacher model” that works only on clean inputs. The features of both the clean and robust teachers are transferred to the student through a cross-layer attention matrix. Experimental results on the CIFAR-10 dataset show that the proposed method improves classification accuracy on clean samples compared to the original IBD, while maintaining similar accuracy on adversarial samples. We also analyze the impact of different training settings that have different influences on the attention module.
Keyword (in Japanese) (See Japanese page) 
(in English) Adversarial Attacks / Adversarial Robustness / Information Bottleneck / Knowledge Distillation / Convolutional Neural Networkd / Image Classification / /  
Reference Info. IEICE Tech. Rep., vol. 125, no. 425, IBISML2025-44, pp. 25-32, March 2026.
Paper # IBISML2025-44 
Date of Issue 2026-03-17 (IBISML) 
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 IBISML2025-44

Conference Information
Committee PRMU IPSJ-CVIM IBISML ITE-SIP  
Conference Date 2026-03-24 - 2026-03-25 
Place (in Japanese) (See Japanese page) 
Place (in English)  
Topics (in Japanese) (See Japanese page) 
Topics (in English)  
Paper Information
Registration To IBISML 
Conference Code 2026-03-PRMU-CVIM-IBISML-SIP 
Language English 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Improving the Robustness/Accuracy Tradeoff in Robust Information Bottleneck Distillation Through Dual Teachers 
Sub Title (in English)  
Keyword(1) Adversarial Attacks  
Keyword(2) Adversarial Robustness  
Keyword(3) Information Bottleneck  
Keyword(4) Knowledge Distillation  
Keyword(5) Convolutional Neural Networkd  
Keyword(6) Image Classification  
Keyword(7)  
Keyword(8)  
1st Author's Name Vincent Ryusuke Takahashi  
1st Author's Affiliation Kyushu University (Kyushu Univ.)
2nd Author's Name Yoshinari Takeishi  
2nd Author's Affiliation Kyushu University (Kyushu Univ.)
3rd Author's Name Jun'ichi Takeuchi  
3rd Author's Affiliation Kyushu University (Kyushu Univ.)
4th Author's Name Kave Salamatian  
4th Author's Affiliation Savoie Mont Blanc University (Savoie Mont Blanc Univ.)
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 2026-03-24 15:20:00 
Presentation Time 20 minutes 
Registration for IBISML 
Paper # IBISML2025-44 
Volume (vol) vol.125 
Number (no) no.425 
Page pp.25-32 
#Pages 8 
Date of Issue 2026-03-17 (IBISML) 

[Return to Top Page]

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