How Neural Networks Factor Problems of Sensory-Motor Control

Abstract

Some recent results in neural networks relevant to sensory-motor control problems are discussed. A common finding in biologically-oriented neural networks research is that many networks operate in parallel to ensure desired operating characteristics. Results on trajectory formation, sensory updating, and anticipatory compensation illustrate networks that are applicable in several performance domains (planned arm and speech movements, ballistic eye-movements) and that help explain data on several distinct but cooperative brain regions (frontal cortex, globus pallidus, cerebellum). In these systems whose fast dynamics are governed by slowly changing network transmission weights as well as by rapidly fluctuating external inputs, a major focus of research is how to ensure that short-term dynamics automatically regulate learning (transmission weight modification) in such a way that the system is guaranteed to develop along an adaptive trajectory.