Adaptive Development of Connectionist Decoders for Complex Error-Correcting Codes

Part of Advances in Neural Information Processing Systems 4 (NIPS 1991)

Bibtex Metadata Paper


Sheri Gish, Mario Blaum


\Ve present. an approach for df'velopment of a decoder for any complex binary code- (ECC) via training from examples of decoded received words. Our decoder is a connectionist architecture. We describe two sepa.rate solutions: A system-level solution (the Cascaded Networks Decoder); and the ECC-Enhanced Decoder, a solution which simplifies the mapping problem which must be solved for decoding. Although both solutions meet our basic approach constraint for simplicity and compact(cid:173) ness. only the ECC-Enhanced Decoder meet.s our second basic constraint of being a generic solution.