Association performance of cross-coupled Hopfield nets for correlated patterns

This paper describes the improvement on the association dynamics of cross-coupled Hopfield nets with many-to-many mapping internetworks (CCHN-MMMI) for correlated patterns. CCHN-MMMI is composed of plural Hopfield-type neural networks which are mutually connected via multilayered networks (internetworks). The association performance of CCHN-MMMI is evaluated for correlated pattern sets through the computer simulations. The simulation results show that the storage capacity of CCHN-MMMI greatly increases as compared with that of Hopfield-type associative memory with pseudo-inverse matrix of weights. Especially, CCHN-MMMI with nonlinear mapping internetworks has useful characteristics for the memory and the retrieval of correlated patterns.

[1]  Masahiko Morita,et al.  Associative memory with nonmonotone dynamics , 1993, Neural Networks.

[2]  Bart Kosko,et al.  Unsupervised learning in noise , 1990, International 1989 Joint Conference on Neural Networks.

[3]  Teuvo Kohonen,et al.  Self-Organization and Associative Memory , 1988 .

[4]  K. Tsutsumi,et al.  Cross-coupled Hopfield nets via generalized-delta-rule-based internetworks , 1990, 1990 IJCNN International Joint Conference on Neural Networks.

[5]  Kosko.  Unsupervised learning in noise , 1989 .

[6]  Shun-ichi Amari,et al.  Characteristics of sparsely encoded associative memory , 1989, Neural Networks.