Indirect differentiation of function for a network of biologically plausible neurons

Abstract

This paper introduces a new method to model differentiation of biologically plausible neurons, introducing the capability for indirectly defining the characteristics for a network of spiking neurons. Due to its biological plausibility and greater potential for computational power, a spiking neuron model is employed as the basic functional unit in our system. The method for designing the architecture (network design, communication structure, and neuron functionality) for networks of spiking neurons has been purely a manual process. In this paper, we propose a new design for the differentiation of a network of spiking neurons, such that these networks can be indirectly specified, thus enabling a method for the automatic creation of a network for a predetermined function. In this manner, the difficulties associated with the manual creation of these networks are overcome, and opportunity is provided for the utilization of these networks more readily for applications. Thus, this paper provides a new method for indirectly constructing these powerful networks, such as could be easily linked to an evolutionary system or other optimization algorithm. © Springer-Verlag Berlin Heidelberg 2003.