Multiple RIS-assisted federated learning system based on channel quality device selection algorithm

Machine learning has been introduced to assist humans in handling complex computations across various domains. However, concerns about inadequate user privacy have hindered its widespread adoption in communication systems. To enhance communication efficiency and reduce the overhead associated with machine learning, reconfigurable intelligent surfaces (RIS) and over-the-air computation (AirComp) have been utilized as supportive technologies. Simultaneously, federated learning (FL) emerged as an alternative approach to preserve user privacy and mitigate communication latency issues by requiring only the exchange of gradient information. Research has demonstrated that integrating FL with RIS leads to improved performance in system models. In this context, we propose a multiple RIS-assisted over-the-air FL system that incorporates a device selection algorithm based on channel quality. This algorithm identifies optimal devices for participation in the training process according to their communication channel conditions. Simulation results confirm that our proposed model outperforms the benchmark designs in terms of test accuracy and convergence speed.