MULTIPLE PERIODIC PATTERNS VIA DISCRETE NEURAL NETS WITH DELAYED FEEDBACK LOOPS

JIANHONG WU, RUYUAN ZHANG, XINGFU ZOU
2004 International Journal of Bifurcation and Chaos in Applied Sciences and Engineering  
We show that a discrete dynamical system with delayed arguments that describe the computational performance of a neural network with a loop structure can exhibit the coexistence of huge number of periodic solutions, and we describe the domains of attraction for the stable periodic orbits. This demonstrates a great potential of discrete time delayed neural nets with a loop structure for the purpose of storage and retrieval of periodic patterns.
doi:10.1142/s0218127404011090 fatcat:5pm6crn3fzel7lxes26wzqys54