Federal learning client selection method and system for long-tail data

• Main Authors: SUN YING, HUANG JIAJIA, WEI FUSHENG
• Format: Patent
• Language: Chinese; English
• Published: 05.01.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC COMMUNICATION TECHNIQUE; ELECTRICITY; PHYSICS; TRANSMISSION; TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHICCOMMUNICATION

The invention provides a federal learning client selection method and system for long-tail data, and the method comprises the steps: calculating the contribution score of each client, and selecting the first N clients with the contribution scores ranked from high to low; local models uploaded by the first N clients are received, one or more clients meeting the uploading time delay requirement serve as selected clients, and the local models are obtained by conducting long-tail data training on a first global model issued by a server through the clients; aggregating the local models uploaded by all the selected clients, and updating the first global model according to the aggregated models to obtain a second global model; according to the method, the influence of long-tail data and a wireless fading channel on the performance of the federated learning model can be reduced, and the universality and robustness of the model are improved. 本发明提供了一种面向长尾数据的联邦学习客户端选择方法及系统，包括：计算每个客户端的贡献得分，并选取贡献得分由高到低排序的前N个客户端；接收所述前N个客户端