Boosting federated learning in optical networks for edge computing

Jun Li; Jiajia Chen

doi:10.1117/12.2582353

5 March 2021 Boosting federated learning in optical networks for edge computing

Jun Li, Jiajia Chen

Proceedings Volume 11712, Metro and Data Center Optical Networks and Short-Reach Links IV; 117120J (2021) https://doi.org/10.1117/12.2582353
Event: SPIE OPTO, 2021, Online Only

Abstract

This invited talk will introduce several useful means to boost federated learning carried out over passive optical networks for edge computing. One way is to introduce bandwidth slicing that is able to reserve network resource dedicated to the learning task and can well address the issues brought by stragglers during the training process. Another way is to devise aggregation function carried out at intermedium network nodes capable of significantly reducing the amount of traffic to be exchanged for global training while not impacting the learning performance. Simulation results show that the FL training efficiency can be significantly improved while achieving the same level of learning accuracy. For the specific FL task implemented in the context of edge computing, the training time can be saved up to 36 % to achieve the maximum learning accuracy.

Conference Presentation

Citation Download Citation

Jun Li and Jiajia Chen "Boosting federated learning in optical networks for edge computing", Proc. SPIE 11712, Metro and Data Center Optical Networks and Short-Reach Links IV, 117120J (5 March 2021); https://doi.org/10.1117/12.2582353

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available