Training Asymptotically Stable Recurrent Neural Networks

Abstract -In this work we present a class of recurrent fashion. For example, the axons of the granule cells become networks which are asymptotically stable. For these networks, elongated and are arranged in parallel to each other forming the we discuss their similarity with certain structures in the central Parallel Fibers. Neurons from all the classes but granule cells, renervous system, and prove that if an interconnection pattern that ceive input from the Parallel Fibers. The absence of loops (e.g. does not allow excitatory feedback is used, then the resulting regranule cells to granule cells) contributes to the stability of the current neural network is stable. We introduce a training methstructure, as we shall prove subsequently. Additionally, the arodology for networks belonging to this class, and use it to train borizations of both the dendrites and the axons are limited, and networks that successfully identify a number nonlinear systems. thus neurons are affected by neurons which are in their immedi-