Neural-Polar code:An inventing channel coding scheme based on deep learning
To address the need for intelligence in new mobile communication systems and the challenge of reliable commu-nication 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 intro-duced 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 intelli-gence in encoding and decoding.
万方数据
维普
