Title :
Robust learning rule for bidirectional associative memory
Author_Institution :
Dept. of Comput. Sci., Chinese Univ. of Hong Kong, Shatin, Hong Kong
Abstract :
A robust learning rule, called adaptive Ho-Kashyap bidirectional learning (AHKBL), is proposed to enhance the capacity and error correction capability of a bidirectional associative memory (BAM). Also, the sufficient conditions for convergence of AHKBL are discussed. Simulation shows that AHKBL greatly improves the capacity and the error correction capability of the BAM.
Keywords :
content-addressable storage; error correction; learning (artificial intelligence); neural nets; adaptive Ho-Kashyap bidirectional learning; bidirectional associative memory; convergence; error correction capability; robust learning rule; sufficient conditions; Associative memory; Computer science; Convergence; Encoding; Error correction; Libraries; Magnesium compounds; Neurons; Robustness; Sufficient conditions;
Conference_Titel :
Neural Networks, 1993. IJCNN '93-Nagoya. Proceedings of 1993 International Joint Conference on
Print_ISBN :
0-7803-1421-2
DOI :
10.1109/IJCNN.1993.714277
