Federated continual learning based on weight self-optimization algorithm in non-iid

Hao Xu

doi:10.1117/12.3032841

5 July 2024 Federated continual learning based on weight self-optimization algorithm in non-iid

Hao Xu

Proceedings Volume 13184, Third International Conference on Electronic Information Engineering and Data Processing (EIEDP 2024); 131843H (2024) https://doi.org/10.1117/12.3032841
Event: 3rd International Conference on Electronic Information Engineering and Data Processing (EIEDP 2024), 2024, Kuala Lumpur, Malaysia

Abstract

Federated Learning enables collaboration among parties without the need for external data sharing, achieving knowledge sharing without data sharing. However, existing federated learning algorithms still encounter issues such as non-IID (nonindependent and identically distributed) data and catastrophic forgetting, such as data transfer, statistical differences, and local update overfitting. In this paper, we propose a new federated learning algorithm, FCLW, which integrates continuous learning to address catastrophic forgetting and introduces a weight auto-allocation mechanism to solve the heterogeneity of federated learning data. Experimental results demonstrate that, compared to baseline algorithms, FCLW achieves faster convergence and obtains a more accurate global model in both IID and non-IID data settings.

(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Hao Xu "Federated continual learning based on weight self-optimization algorithm in non-iid", Proc. SPIE 13184, Third International Conference on Electronic Information Engineering and Data Processing (EIEDP 2024), 131843H (5 July 2024); https://doi.org/10.1117/12.3032841

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