An Approximation to Compartmental Neuronal Modelling with Significant Gains in Performance

In this paper I present the results of an investigation into modelling neurons by a method that approximates the realistic dynamics achieved by compartmental modelling but with significantly less computation. The method employs artificial neural networks as the components for modelling voltage- and time-dependent conductances, and lengths of membrane. I describe a model of a squid giant axon and indicate the computational gains over a comparable compartmental model. A generalized set of such artificial neural networks could provide the means to conduct fast, largescale simulations. keywords: neural networks, computational neuroscience, compartmental modelling Introduction Computational neuroscience is a rapidly evolving field of research which is chiefly concerned with building mathematical models and computer simulations of nervous systems. It has proven invaluable in guiding experimentation [Sejnowski, Koch, and Churchland, 1990], understanding complex neurodynamics [Traub, Wong, M..