A perspective on research aimed at understanding the systems nature of neural controllers

Summary form only given, as follows. In biological systems, motor control as well as other information processing tasks are carried out by networks of neural processors. Over the years, researchers have attained considerable understanding of the use of stochastic algorithmic approaches toward the control of physiological systems. In distinction to that, the authors aim at understanding how neural networks themselves do adaptively and incrementally learn to implement effective control action. They place their work in perspective relative to the adaptive stochastic control approach on one hand and to the completely heuristic reward/punish approaches on the other. Related work of the same nature is also discussed. Emphasis is on understanding the research issues and the approach, but illustrative results are also presented.<>