Optimization Algorithm for Cloud-Edge-Terminal Federated Learning in the IoT

With the extensive application of the Internet of Things (IoT) in multiple fields, diversified scenarios put forward strict and differentiated requirements for quality of service. The traditional centralized cloud computing architecture is difficult to meet the needs of large-scale data processing. Although federated learning is a potential solution for model training in cloud-edgeend architectures, it faces challenges such as non-IID data and differences in device computing resources. Moreover, IoT systems have the coupling problem of security constraints and delay optimization. To this end, this paper designs an adaptive weight aggregation mechanism to dynamically sense data distribution and adjust gradient aggregation weights by combining the contribution of device computing resources, so as to improve the convergence speed and adaptability of the model in non-IID environments; ADMM is introduced to solve the coupling problem of security and delay through cross-layer collaborative solving. Comparative experiments on the MNIST dataset show that the algorithm performs better in terms of accuracy, precision, recall rate, and F1 score and other indicators.