Sparse code multiple access (SCMA) has become a highly competitive technology for future cellular systems. For the receiver of the SCMA system, besides the traditional maximum likelihood and message passing algorithm solutions, a deep neural network (DNN) method that causes whirlwinds in image recognition can reduce the computational complexity of the decoder. We expect low complexity while maintaining a satisfactory bit error rate (BER) performance. As shown in our simulations, our proposed solution has better BER performance and lower computational complexity than other previously studied DNN solutions.

中文翻译：

---

通过深度学习进行SCMA解码

稀疏代码多路访问（SCMA）已成为未来蜂窝系统的高度竞争技术。对于SCMA系统的接收器，除了传统的最大似然和消息传递算法解决方案之外，在图像识别中引起旋风的深度神经网络（DNN）方法还可以降低解码器的计算复杂度。我们期望低复杂度，同时保持令人满意的误码率（BER）性能。如我们的仿真所示，与其他先前研究的DNN解决方案相比，我们提出的解决方案具有更好的BER性能和更低的计算复杂度。