<![CDATA[Federated Learning is a promising communication model to address data security and privacy issues. Each client device engages in a collaborative machine learning model, eliminating the need to send all client data to the server. However, the main obstacles to applying FL to wireless network communication are limited bandwidth and unstable network conditions. Therefore, this research proposes a new FedFA approach integrating the Firefly algorithm to optimize weight initialization and minimize communication costs. The basic principle of FedFA involves parameters in the Firefly algorithm to select the best weight of each client to be trained on the server. Based on the test results, the proposed algorithm produces an accuracy improvement of 12.84% compared to FedAvg. The FedFA model is also more resilient to unstable communication, as seen from the less significant decrease in accuracy compared to the FedAvg algorithm.]]>
Copyrights © 2025
