Deep Neural Network Based Decoder Design for 5G Codes
Project Period: 2018–2019
Customer: Huawei Technologies Co. Ltd.
Project leader: A. Frolov
Description: The main goal is to suggest an application of deep learning to channel decoding in order to improve performance or reduce complexity/latency. There were already attempts to construct neural network (NN) learning-based decoders in literature, here we face with so-called curse of dimensionality problem: even for a short code of length N = 100 bits and rate R =0.5, 250 different codewords exist, which are far too many to fully train any NN in practice. In our opinion, the only way deal with the problem is to combine deep learning methods with existing decoding algorithms.
