Associative Memory by Using Coupled Gaussian Maps

The associative memory model comprised of coupled Gaussian maps is proposed. The Gaussian map is a one-dimensional discrete-time dynamical system, which generates various phenomena including periodic and non-periodic points. The Gaussian associative memory has similar characteristics of both Hopfield and chaos neural associative memories, and it can change those modes by just changing the system parameters. When the Gaussian associative memory successively recalls the stored patterns in such manner as the chaotic associative memory, the Gaussian associative memory also recalls some pseudo patterns which were not actually stored into the memory. It was found that the pseudo patterns corresponded to the chaotic trajectories generated in the Gaussian associative memory. Therefore, by using the method of avoiding chaotic behavior, we could eliminate the generation of the pseudo patterns. In this paper, we introduce the dynamics of the Gaussian associative memory model and present the simulation results. In addition, the output patterns obtained by the Gaussian associative memory with/without the function of avoiding chaos are presented.