In a global neural network, which consists of many connected local neural networks the mechanism of memory and the information flow are discussed. One of local neural networks in the global network is connected with an input network, which provides some pattern (or some information) to the global network from outside. We assume the nearest-neighbor coupling between local networks in the global netowrk. The local netowrk can exchange the pattern or information among nearest-neighbor local networks in the global network. The local netowrk can exchange the pattern or information among nearest-neighbor local netowrks and the patterns can propagate in the global system. The propagation of the pattern persists even after the connection between the input network and one of local networks in the global network was decoupled. For the appropriate nearest-neighbor coupling the propagated pattern generates a global pattern and it does not disappear. We study the role of the local coupling in the genesis of a global pattern and in memorization of the information inputted from the outside.
"Mechanism of memory in connected neural networks", Proc. SPIE 5851, Fundamental Problems of Optoelectronics and Microelectronics II, (8 June 2005); doi: 10.1117/12.634384; https://doi.org/10.1117/12.634384