GBPNet neural decoder improves linear code performance over belief propagation
Original title: Learning to Decode Linear Codes with Generalized Belief Propagation Network
GBPNet preserves Tanner-graph symmetry with cascaded variable-node updates and learned message scaling to achieve lower bit error rates than conventional belief propagation.
Error-correcting codes enable reliable transmission over noisy channels. Neural decoders have shown gains over classical iterative methods. GBPNet uses cascaded variable-node updates and element-wise message scaling while retaining the Tanner graph structure. Progressive and high-SNR training strategies improve accuracy. Experiments on several linear codes confirm lower bit error rates than belief propagation and prior neural decoders.
- Cascaded variable-node updates retain message-passing symmetry
- Learned element-wise message scaling enhances decoder capability
- Two training strategies: progressive and high-SNR
- Lower bit error rates than conventional belief propagation
- Evaluated on several linear codes
From the original article
1 min read at arxiv.orgError correcting codes are a fundamental component of communication systems, enabling reliable data transmission over noisy channels. Recently, neural decoders have demonstrated notable performance gains over classical iterative algorithms. In this paper, we propose a Generalized Belief Propagation Network (GBPNet) that preserves message-passing symmetry while enhancing the decoder's learning capability.
By Samira Malek · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.