Modeling Neuronal Systems

Abstract

This chapter provides a summary of current approaches to modeling neuronal systems at the levels of single cells, networks, and more complex multinetwork systems. It begins with a brief history describing how models based on neurophysiological data diverged from artificial intelligence research and became increasingly sophisticated as available computational power increased. It is shown how, in order to make simulation of large network systems practical, models based on detailed ion channel properties can be replaced by increasingly simple “integrate-and-fire” models, “rate-coded” models that do not instantiate individual neuronal action potential spikes, and even ensemble models that provide statistical summaries of the activity of masses of neurons. Event-driven models that reduce the need to perform routine membrane-potential decay calculations at small time intervals are discussed. Models for synaptic modification presumed to be involved in learning are described for both rate-coded and spiking neuron models. The chapter ends with some discussion of nervous system aspects that are often omitted from models and which may form suitable bases for future research, as well as pitfalls that need to be avoided by newcomers to this field.