Convolutional Neural Network Based Optimization Approach for Wireless Resource Management
Md. Habibur Rahman, Md Munjure Mowla, Shahriar Shanto · 2020
In this paper, the feasibility of evolving advanced deep learning technology is demonstrated to solve the NP-hard transmit power control problem for future wireless networks. In the existing deep learning approach, the larger network size adversely affects the performance of the considered fully connected multi-layer perceptron (MLP) network. A Convolutional Neural Network (CNN) based strategy is proposed to develop a scalable method for solving high computation time of conventional power control algorithms. The findings reveal that the sum-rate performance of the proposed CNN outperforms existing neural network based strategy and speeds up the operation to a significant rate over the typical power control algorithm, namely Stochastic Weighted Minimum Mean Square Error (SWMMSE). The proposed scheme offers an improved solution to the optimization algorithm's high computation time problem of radio resource allocation in real-time applications.