Deep Neural Network based Channel Allocation for Interference-Limited Wireless Networks

Cooperative communication in wireless networks has received much attention in both academia and industry. How to effectively allocate and schedule radio resources to improve system performance becomes an important issue of cooperative communication. This paper mainly studies the ultra-low complexity wireless channel allocation algorithm for interference-limited networks. Firstly, we use the traditional sequential convex approximation (SCA) technique to design the channel allocation algorithm. Then, we utilize the characteristics of deep neural network (DNN) that can approximate a complex function with multiple layers of mapping to approximate the SCA-based algorithm. Based on DNN, we design an ultra-low complexity algorithm. Simulation results indicate that the DNN-based algorithm can achieve good performance with ultra-low computation time, which is a feature for practical application.