Federated Learning Technology in Serial Topology for IoT Networks

Recently, with the rapid development of IoT technology, a new type of distributed collaborative AI technology, federated learning is used on the IoT platform to solve the problems of big data and privacy protection. However, federated learning still has a long way from being practical as many key technologies still need to be broken through. This paper considers a federated learning IoT system under serial topology to solve resource consumption problems such as communication. This paper mainly studies algorithm optimization strategies including the impact of different federation aggregation methods, the balance between communication resources and computing power, and the influence of multi-level central nodes on system performance. The results show that the accuracy weighting is the optimal federation aggregation mode, and the algorithm convergence will be slower in most cases because of reducing communication consumption even if the computing power is increased.