Doubly Residual Neural Decoder: Towards Low-Complexity High-Performance Channel Decoding

Recently deep neural networks have been successfully applied in channel coding to improve the decoding performance. However, state-of-the-art decoders cannot achieve high performance and low complexity simultaneously. To overcome this challenge, paper we propose doubly residual (DRN) decoder. By integrating both input learning design of decoder, DRN enables significant improvement while maintaining complexity. Extensive experiment results show that on different types codes, our decoder consistently outperform terms performance, model sizes computational cost.