Aspects of reconfigurable neural networks

摘要

A basic model of a neural processing system is described, inspired by the structure and functional characteristics of the reticular activating system in the brain. Through the capability of the model to change its processing conditions rapidly, the phenomenon of reconfigurability and the usual activity mode of neural networks are related. Several variants of the basic model are considered, as well as their dynamic properties, stability, and fault-tolerance. The applicability of these models is generally discussed with emphasis on the possibility of the creation of complex processing systems.