On partial neural-net decoding with retransmissions for data communications

The authors discuss basic issues on partial decoding of linear block codes with retransmissions by recently developed neural network technology. In particular, the basic structure of a neural-net decoder is defined based on syndrome decoding methods. Partial decoding by the pocket algorithm and the random sampling algorithm is proposed, and its throughput performance is examined. The complexity issues of using feedforward neural networks for decoding programs are also addressed. Neural-net decoding could be a good alternative for providing an error control mechanism for data communications