Title :
Three-layer bidirectional asymmetrical associative memory
Author_Institution :
Dept. of Electr. & Comput. Eng., Windsor Univ., Ont., Canada
Abstract :
In this paper, a recurrent-type of neural network in the form of a three-layer bidirectional symmetrical or asymmetrical associative memory is presented. The network possesses desired structural features of a bidirectional associative memory and a three-layer feedforward neural network. The network allows two modes of memory recall, namely, recalling by a pattern-pair from both layers 0 and 2 inputs, and recalling by one pattern from either the layer 0 or 2 input. Simulation results on noisy alpha-numeral type of autoassociative and heteroassociative pattern recognition problems are given.
Keywords :
content-addressable storage; feedforward neural nets; learning (artificial intelligence); pattern recognition; recurrent neural nets; asymmetrical associative memory; autoassociative pattern recognition problems; heteroassociative pattern recognition problems; learning algorithm; memory recall modes; recurrent-type neural network; symmetrical associative memory; three-layer bidirectional associative memory; three-layer feedforward neural network; Associative memory; Convergence; Error correction; Feedforward neural networks; Magnesium compounds; Multi-layer neural network; Neural networks; Pattern recognition; Recurrent neural networks; Vectors;
Conference_Titel :
Circuits and Systems, 2003. ISCAS '03. Proceedings of the 2003 International Symposium on
Print_ISBN :
0-7803-7761-3
DOI :
10.1109/ISCAS.2003.1205061
