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 2022-01-24 16:20
Addition of DPU Training Function by Tail Layer Training
Yuki Takashima, Akira Jinguji, Hiroki Nakahara (Tokyo Tech) VLD2021-59 CPSY2021-28 RECONF2021-67
Abstract (in Japanese) (See Japanese page) 
(in English) The demand for deep learning has been increasing, and many hardware implementations have been proposed. The Deep learning Processor Unit (DPU) was provided by Xilinx. Althogh it can perform inference at high speed, it cannot perform training.
We propose a tail layer training that makes the tail layer of a Convolutional Neural Network (CNN) independent. All layers except the tail layer are computed using a DPU, and the tail layer is computed by a CPU. Since the number of neurons and classes in the output must be the same for image classification, it is effective for retraining to add the number of classes.
The tail layer training, found that the relationship between the existing classes and the classes to be added is important. Therefore, it is not suitable for training on large number of classes. However, with a dataset such as cifar10, it is able to reduce the loss of accuracy by about 3 points between training the entire model with all 10 classes and training only the tail layer with 2 add classes after training the entire model with 8 classes.
Keyword (in Japanese) (See Japanese page) 
(in English) CNN / Image Classification / DPU / Tail Layer Training / / / /  
Reference Info. IEICE Tech. Rep., vol. 121, no. 344, RECONF2021-67, pp. 55-60, Jan. 2022.
Paper # RECONF2021-67 
Date of Issue 2022-01-17 (VLD, CPSY, RECONF) 
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 VLD2021-59 CPSY2021-28 RECONF2021-67

Conference Information
Committee RECONF VLD CPSY IPSJ-ARC IPSJ-SLDM  
Conference Date 2022-01-24 - 2022-01-25 
Place (in Japanese) (See Japanese page) 
Place (in English) Online 
Topics (in Japanese) (See Japanese page) 
Topics (in English) FPGA Applications, etc. 
Paper Information
Registration To RECONF 
Conference Code 2022-01-RECONF-VLD-CPSY-ARC-SLDM 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Addition of DPU Training Function by Tail Layer Training 
Sub Title (in English)  
Keyword(1) CNN  
Keyword(2) Image Classification  
Keyword(3) DPU  
Keyword(4) Tail Layer Training  
Keyword(5)  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Yuki Takashima  
1st Author's Affiliation Tokyo Institute of Technology (Tokyo Tech)
2nd Author's Name Akira Jinguji  
2nd Author's Affiliation Tokyo Institute of Technology (Tokyo Tech)
3rd Author's Name Hiroki Nakahara  
3rd Author's Affiliation Tokyo Institute of Technology (Tokyo Tech)
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 2022-01-24 16:20:00 
Presentation Time 25 minutes 
Registration for RECONF 
Paper # VLD2021-59, CPSY2021-28, RECONF2021-67 
Volume (vol) vol.121 
Number (no) no.342(VLD), no.343(CPSY), no.344(RECONF) 
Page pp.55-60 
#Pages 6 
Date of Issue 2022-01-17 (VLD, CPSY, RECONF) 

[Return to Top Page]

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