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 2026-06-12 14:25
On the Impact of Residual Connections on Hierarchical Feature Learning in CNNs
Jirayus Lapamnuaypol, Kenya Jin'no (TCU Univ.) NLP2026-23 CCS2026-23
Abstract (in Japanese) (See Japanese page) 
(in English) Residual connections are a fundamental component of modern convolutional neural networks, enabling the successful optimization of deep architectures. While their benefits for gradient propagation and optimization are well established, their influence on hierarchical feature learning remains less explored. This study investigates how residual learning affects internal representation development in convolutional networks from two complementary perspectives. First, t-SNE visualization is applied to intermediate feature representations in ResNet-50 to examine layer-wise feature evolution and residual block behavior. The analysis reveals that while hierarchical abstraction is preserved globally, residual addition partially restores earlier feature manifold structures, indicating that skip connections modify strictly sequential refinement. Second, a dynamic activation function, Activity-Modulated ReLU (AMReLU), is introduced to probe layer-wise activity propagation through adaptive threshold modulation based on preceding-layer activations. Experiments on plain CNNs and ResNet-20 models trained on CIFAR-10 show that AMReLU produces structured sparsity dynamics in sequential architectures, while these dynamics become attenuated in residual networks due to feature mixing introduced by skip connections. These findings suggest that residual learning preserves intermediate representations across depth, resulting in a more iterative hierarchical refinement process compared to conventional feedforward CNNs.
Keyword (in Japanese) (See Japanese page) 
(in English) Residual connection / t-SNE / feature Representation / / / / /  
Reference Info. IEICE Tech. Rep., vol. 126, no. 68, NLP2026-23, pp. 119-124, June 2026.
Paper # NLP2026-23 
Date of Issue 2026-06-04 (NLP, CCS) 
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 NLP2026-23 CCS2026-23

Conference Information
Committee CCS NLP  
Conference Date 2026-06-11 - 2026-06-12 
Place (in Japanese) (See Japanese page) 
Place (in English) I-site Namba 
Topics (in Japanese) (See Japanese page) 
Topics (in English) Nonlinear Problems, Complex Communication Sciences, etc. 
Paper Information
Registration To NLP 
Conference Code 2026-06-CCS-NLP 
Language English 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) On the Impact of Residual Connections on Hierarchical Feature Learning in CNNs 
Sub Title (in English)  
Keyword(1) Residual connection  
Keyword(2) t-SNE  
Keyword(3) feature Representation  
Keyword(4)  
Keyword(5)  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Jirayus Lapamnuaypol  
1st Author's Affiliation Graduate School of Tokyo City University (TCU Univ.)
2nd Author's Name Kenya Jin'no  
2nd Author's Affiliation Graduate School of Tokyo City University (TCU Univ.)
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 ()
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 2026-06-12 14:25:00 
Presentation Time 25 minutes 
Registration for NLP 
Paper # NLP2026-23, CCS2026-23 
Volume (vol) vol.126 
Number (no) no.68(NLP), no.69(CCS) 
Page pp.119-124 
#Pages 6 
Date of Issue 2026-06-04 (NLP, CCS) 

[Return to Top Page]

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