Asymmetric neural networks incorporating the Dale hypothesis and noise-driven chaos

Abstract

Dynamical properties of the neural networks with asymmetrical synaptic couplings respecting the Dale hypothesis are studied. The time evolution of the networks is assumed to obey stochastic dynamics of the Little type with time delay. Using a nonlinear master equation, exact equations are derived for the time evolution of the overlaps of instantaneous configuration with p embedded patterns and with the characteristic pattern representing the configuration of excitatory and inhibitory neurons. It is shown that the networks exhibit noise-driven chaotic motions in the retrieval process.