2D Mesh Image Subcarrier Assignment for MUDiv/OFDMA with Convolutional Neural Networks

In an orthogonal frequency division multiple access (OFDMA), the subcarrier assignment with the multiuser diversity (MUDiv) gain has been proposed. However, in a MUDiv/OFDMA, the large computational complexity is required to obtain the good system performance. To solve this problem, convolutional neural networks (CNN) is adopted. CNN can contribute to reducing the computational complexity. However, when the learning data is not suitable, the MUDiv gain is not obtained, and the system performance is deteriorated. Therefore, in this paper, we propose the 2 dimensional (2D) mesh image subcarrier assignment for the MUDiv/OFDMA with the CNN.