Energy-Efficient Mutual Learning Over D2D Communications

Mutual learning (ML) is a promising technique for sharing knowledge in data while keeping the data privacy and preserving the individual characteristics of the local model. In this article, we design a novel decentralized mutual learning (DML) system, where bidirectional device-to-device (D2D) communications are employed to facilitate the knowledge sharing. To accelerate the learning process and reduce the energy consumption at mobile devices, a non-convex optimization problem is formulated to minimize the average communication energy consumption for sharing knowledge among mobile devices. On this basis, a two-layer iterative algorithm is proposed, which consists of an outer layer algorithm based on the particle swarm optimization (PSO) method for searching a suitable user selection strategy and an inner layer algorithm based on sum-of-ratios optimization method to achieve a globally optimal allocation of communication resource for accelerating the learning process. Numerical results validate the fast convergence property and the effectiveness of the proposed algorithm, and the asynchronous nature of the designed system in terms of knowledge sharing and energy saving.