| Paper Abstract and Keywords |
| Presentation |
2026-01-28 16:15
A Proposal of Dymamic Optimization for Vision Transformer Hongyi Zhang, Yukari Yamauchi (Nihon Univ.) NLP2025-95 MBE2025-35 NC2025-57 |
| Abstract |
(in Japanese) |
(See Japanese page) |
| (in English) |
In this study, we focus on the Vision Transformer (ViT), which extends the Transformer architecture—originally achieving breakthrough results in natural language processing—to the image domain. While ViT is effective for tasks such as image classification, its performance is highly dependent on large-scale training data and tends to degrade when trained on small datasets. To improve the computational efficiency of Vision Transformers, this study proposes a hybrid architecture that combines Gated Positional Self-Attention (GPSA) with dynamic token pruning based on Proximal Policy Optimization (PPO). Through experiments on CIFAR-10 at 224×224 resolution, we aim to achieve substantial FLOPs reduction while maintaining accuracy by jointly learning the gating mechanism and pruning ratio. |
| Keyword |
(in Japanese) |
(See Japanese page) |
| (in English) |
ViT / Small Datasets / Reinforcement Learning / / / / / |
| Reference Info. |
IEICE Tech. Rep., vol. 125, no. 346, NC2025-57, pp. 53-57, Jan. 2026. |
| Paper # |
NC2025-57 |
| Date of Issue |
2026-01-21 (NLP, MBE, 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 |
NLP2025-95 MBE2025-35 NC2025-57 |
| Conference Information |
| Committee |
NC MBE NLP IEE-MBE |
| Conference Date |
2026-01-28 - 2026-01-30 |
| Place (in Japanese) |
(See Japanese page) |
| Place (in English) |
Kyushu Institute of Technology, Wakamatsu Campus |
| Topics (in Japanese) |
(See Japanese page) |
| Topics (in English) |
NC, NLP, ME, etc. |
| Paper Information |
| Registration To |
NC |
| Conference Code |
2026-01-NC-MBE-NLP-MBE |
| Language |
Japanese |
| Title (in Japanese) |
(See Japanese page) |
| Sub Title (in Japanese) |
(See Japanese page) |
| Title (in English) |
A Proposal of Dymamic Optimization for Vision Transformer |
| Sub Title (in English) |
|
| Keyword(1) |
ViT |
| Keyword(2) |
Small Datasets |
| Keyword(3) |
Reinforcement Learning |
| Keyword(4) |
|
| Keyword(5) |
|
| Keyword(6) |
|
| Keyword(7) |
|
| Keyword(8) |
|
| 1st Author's Name |
Hongyi Zhang |
| 1st Author's Affiliation |
Graduate School of Nihon University (Nihon Univ.) |
| 2nd Author's Name |
Yukari Yamauchi |
| 2nd Author's Affiliation |
Nihon University (Nihon 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 |
2026-01-28 16:15:00 |
| Presentation Time |
25 minutes |
| Registration for |
NC |
| Paper # |
NLP2025-95, MBE2025-35, NC2025-57 |
| Volume (vol) |
vol.125 |
| Number (no) |
no.344(NLP), no.345(MBE), no.346(NC) |
| Page |
pp.53-57 |
| #Pages |
5 |
| Date of Issue |
2026-01-21 (NLP, MBE, NC) |