Belief Propagation
An inference algorithm that performs iterative message passing on factor graphs such as Tanner graphs. It is widely used for linear code decoding, achieving decoding by alternately passing posterior and check messages between variable nodes and check nodes.
Timeline (last 90 days)
The belief propagation (BP) algorithm performs iterative message passing on the Tanner graph and is the mainstream method for linear code decoding, but under certain code types and channel conditions its convergence and decoding accuracy are unsatisfactory.
BP decoding alternates iteratively on the Tanner graph between variable nodes sending posterior information to check nodes and check nodes sending check information back to variable nodes.
The sparsity of the Tanner graph determines the computational complexity of BP, while the presence of short cycles (girth) in the graph undermines the independence assumption of messages and thereby causes convergence difficulties.
Experiments on multiple linear codes show that GBPNet achieves a lower bit error rate (BER) than traditional belief propagation and the evaluated state-of-the-art neural decoders.
All Facts (4)
The belief propagation (BP) algorithm performs iterative message passing on the Tanner graph and is the mainstream method for linear code decoding, but under certain code types and channel conditions its convergence and decoding accuracy are unsatisfactory.
50%UnverifiedBP decoding alternates iteratively on the Tanner graph between variable nodes sending posterior information to check nodes and check nodes sending check information back to variable nodes.
50%UnverifiedThe sparsity of the Tanner graph determines the computational complexity of BP, while the presence of short cycles (girth) in the graph undermines the independence assumption of messages and thereby causes convergence difficulties.
50%UnverifiedExperiments on multiple linear codes show that GBPNet achieves a lower bit error rate (BER) than traditional belief propagation and the evaluated state-of-the-art neural decoders.
50%