Neuroethology, computational

Abstract

Over the past decade, a number of neural network researchers have used the term computational neuroethology to describe a specific approach to neuroethology. Neuroethology is the study of the neural mechanisms underlying the generation of behavior in animals, and hence it lies at the intersection of neuroscience (the study of nervous systems) and ethology (the study of animal behavior); for an introduction to neuroethology, see Simmons and Young (1999). The definition of computational neuroethology is very similar, but is not quite so dependent on studying animals: animals just happen to be biological autonomous agents. But there are also non-biological autonomous agents such as some types of robots, and some types of simulated embodied agents operating in virtual worlds. In this context, autonomous agents are self-governing entities capable of operating (i.e., coordinating perception and action) for extended periods of time in environments that are complex, uncertain, and dynamic. Thus, computational neuroethology can be characterised as the attempt to analyze the computational principles underlying the generation of behavior in animals and in artificial autonomous agents.