论文标题
使用深神经网络优化无细胞大型MIMO系统中的光谱效率
Optimization of Spectral Efficiency in Cell-Free massive MIMO Systems Using Deep Neural Networks
论文作者
论文摘要
蜂窝通信是世界上广泛使用的技术，其中覆盖面积分为多个单元。干扰是蜂窝网络中最重要的挑战之一，它通过降低服务质量而导致问题。无细胞（CF）大量多输入多输出（MIMO）是一种新颖的技术，其中大量的分布式访问点（AP）同时提供少量的用户设备（UE）。 CF网络可能是用于减少干扰的蜂窝网络的替代技术。在CF网络中，一个具有挑战性的任务是可伸缩性，尽管UE的数量倾向于无穷大，但计算复杂性必须在每个AP或UE中保持有限。在本文中，我们提供了两个密集的完全连接的神经网络（密度_net）和1D卷积神经网络（CORV_NET）的体系结构，该架构将在每个AP/UE中的天线数量以及组合向量的方法中实施。在所有情况下，dense_net均优于conv_net。例如，在第一种情况下，它的损失方面有所提高％62.87。结果表明，我们提出的方法在获得光谱效率的最佳值方面表现更好（SE）。
Cellular communication is a widely used technology in the world where the coverage area is divided into multiple cells. Interference is one of the most important challenges in cellular networks which causes problems by reducing the quality of the service. Cell-Free (CF) massive multiple-input multiple-output (MIMO) is a novel technolgy in which a large number of distributed access points (APs) are concurrently serving a small number of user equipment (UE). CF network can be an alternative technology to cellular networks for reducing interference. A challenging task in a CF network is scalability, where although the number of UEs tends to infinity, the computational complexity must remain finite in each AP or UE. In this paper, we provide two architectures of Dense fully connected neural network (Dense_Net) and 1D convolution neural network (Conv_Net) to be implemented in different cases in terms of the number of antennas in each AP/UE and the method for the combining vector. The Dense_Net outperforms the Conv_Net in all the cases. For instance, in the first case it has a %62.87 improvement in terms of loss. The results show that our proposed method performs better in terms of obtaining optimal values for spectral efficiency (SE).
