User Selection Aware Joint Radio-and-Computing Resource Allocation for Federated Edge Learning

Edge intelligence refers to utilize a large number of distributed data and computing resources to learn and inference directly at network edge. Federated edge learning (FEEL) coordinates local model training of edge devices and global model aggregation at a server through wireless connection. To speed up the learning and reduce the resource consumption of the network, we formulate a problem of joint transmission time allocation, computing frequency control and user selection. We propose an efficient solution to solve the non-convex problem. Experiments show that the proposed scheme not only accelerates the learning process but also significantly reduce network cost.