这些文件将在两用户非正交多址（NOMA）系统中实现用于信号检测的深度学习方法[1]。三个主要脚本分别是生成训练数据、训练神经网络和生成测试结果。该神经网络用于相位衰落的静态标量信道，用于NOMA系统中同时检测2个用户的单个子载波上的传输符号。考虑并测试了两种情况：一种是导频符号数目较少，另一种是循环前缀长度较短。在这两种情况下，深度学习方法都比传统的信道估计方法具有更强的鲁棒性。(These files are to implement the deep learning method for signal detection in a two-user non-orthogonal multiple access (NOMA) system [1]. The 3 main scr ipts are to generate training data, to train the neural network and to produce testing results, respectively. The neural network is trained for a static scalar channel with phase fading and is used to detect transmitted symbols on a single subcarrier for 2 users simultaneously in a NOMA system. Two scenarios are considered and tested: one is with fewer number of pilot symbols and the other is with shorter length of cyclic prefix. The deep learning method is shown to be more robust than conventional channel estimation methods in both cases.)