Federated Learning for Resource Allocation in Maritime IoT Networks with Compressed Sensing

Maritime transportation and shipping are vital components of global economic development.In order to enhance shipping efficiency and ensure safety, a distributed machine learning architecture can be utilized to achieve it. Due to concerns about data privacy, many shipping data are not shared. Federated learning $(\mathbf{F L})$, as a form of distributed machine learning, offers the advantage of preserving privacy. The problem of communication efficiency and resource joint optimization in FL over communication networks has been studied, and proposed a communication resource allocation algorithm with compressed sensing. Through experiments, we found that compared to the basic FL method with uniform resource allocation, the proposed RACS algorithm significantly improves the communication efficiency of wireless IoT network FL, as well as the learning performance in terms of training loss and testing accuracy.