| 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 | 2023-06-30 10:45 Beyond Exponential Graph: Communication-Efficient Topologies for Decentralized Learning via Finite-time Convergence Yuki Takezawa, Ryoma Sato, Han Bao (Kyoto Univ./OIST), Kenta Niwa (NTT CS Lab.), Makoto Yamada (OIST)  NC2023-14 IBISML2023-14 |
|
| Abstract | (in Japanese) | (See Japanese page) |
| (in English) | Decentralized learning has recently been attracting increasing attention for its applications in parallel computation and privacy preservation. Many recent studies stated that the underlying network topology with a faster consensus rate leads to a better convergence rate and accuracy for decentralized learning. However, a topology with a fast consensus rate, e.g., the exponential graph, generally has a large maximum degree, which incurs significant communication costs. Thus, seeking topologies with both a fast consensus rate and small maximum degree is important. In this study, we propose a novel topology, the Base-(k + 1) Graph, which endows Decentralized SGD with both a faster convergence rate and more communication efficiency than the exponential graph. | |
| Keyword | (in Japanese) | (See Japanese page) |
| (in English) | Decentralized Learning / Federated Learning / Network Topology / Graph / / / / | |
| Reference Info. | IEICE Tech. Rep., vol. 123, no. 91, IBISML2023-14, pp. 83-90, June 2023. | |
| Paper # | IBISML2023-14 | |
| Date of Issue | 2023-06-22 (NC, 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 | NC2023-14 IBISML2023-14 |
|
| Conference Information | |
| Committee | NC IBISML IPSJ-BIO IPSJ-MPS |
| Conference Date | 2023-06-29 - 2023-07-01 |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | OIST Conference Center |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Paper Information | |
| Registration To | IBISML |
| Conference Code | 2023-06-NC-IBISML-BIO-MPS |
| Language | Japanese |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Beyond Exponential Graph: Communication-Efficient Topologies for Decentralized Learning via Finite-time Convergence |
| Sub Title (in English) | |
| Keyword(1) | Decentralized Learning |
| Keyword(2) | Federated Learning |
| Keyword(3) | Network Topology |
| Keyword(4) | Graph |
| Keyword(5) | |
| Keyword(6) | |
| Keyword(7) | |
| Keyword(8) | |
| 1st Author's Name | Yuki Takezawa |
| 1st Author's Affiliation | Kyoto University/OIST (Kyoto Univ./OIST) |
| 2nd Author's Name | Ryoma Sato |
| 2nd Author's Affiliation | Kyoto University/OIST (Kyoto Univ./OIST) |
| 3rd Author's Name | Han Bao |
| 3rd Author's Affiliation | Kyoto University/OIST (Kyoto Univ./OIST) |
| 4th Author's Name | Kenta Niwa |
| 4th Author's Affiliation | NTT Communication Science Laboratries (NTT CS Lab.) |
| 5th Author's Name | Makoto Yamada |
| 5th Author's Affiliation | OIST (OIST) |
| 6th Author's Name | |
| 6th Author's Affiliation | () |
| Speaker | Author-1 |
| Date Time | 2023-06-30 10:45:00 |
| Presentation Time | 25 minutes |
| Registration for | IBISML |
| Paper # | NC2023-14, IBISML2023-14 |
| Volume (vol) | vol.123 |
| Number (no) | no.90(NC), no.91(IBISML) |
| Page | pp.83-90 |
| #Pages | 8 |
| Date of Issue | 2023-06-22 (NC, IBISML) |
[Return to Top Page]