Neural-Polar码:一种基于深度学习的新型信道编码方案

Neural-Polar code:An inventing channel coding scheme basedon deep learning

为应对新型移动通信系统智能性的需求以及在难以进行人工建模的复杂信道环境下进行可靠通信的问题,基于Polar码的编译码递归结构提出一种新型神经网络信道编码方案,即Neural-Polar码。该方案利用神经网络将Polar码编译码递归结构中父、子节点间的线性映射变成非线性映射,引入快速连续抵消(successive cancellation, SC)译码的思想,解决在完全二叉树上构建Neural-Polar码造成网络结构过大的问题。仿真实验表明,Neural-Polar码可以获得优于经典SC译码算法的误码率(bit error rate, BER)和误块率(block error rate, BLER)性能,对网络的联合训练使得Neural-Polar码能够自动学习信道特性,具有更好的信道适应性和鲁棒性。Neural-Polar码将传统的对复杂信道进行人工建模分析的难题交给机器,充分体现出其编译码的智能性。

To address the need for intelligence in new mobile communication systems and the challenge of reliable communication in complex channel environments where manual modeling is difficult,we propose a new neural network channel coding scheme,known as Neural-Polar code,based on the recursive structure of Polar code’s encoding and decoding.In this scheme,the linear mapping between parent and child nodes in the recursive structure of polar coding and decoding is changed into nonlinear mapping by neural network,and the idea of fast successive cancellation decoding(SC)is introduced to solve the problem of too large network structure caused by constructing Neural-Polar codes on complete binary trees.Simulation experiments show that Neural-Polar code can obtain better bit error rate(BER)and block error rate(BLER)than classical successive cancellation(SC)decoding algorithm.Joint training of the network enables Neural-Polar code to automatically learn channel characteristics,resulting in better channel adaptability and robustness.Neural-Polar code transfers the traditional challenge of manually modeling complex channels to machines,fully demonstrating its intelligence in encoding and decoding.