THEORY OF PROBABILITY STATE VARIABLE SYSTEMSnVOLUME VI: PERCEPTION, DECISION-MAKING, AND ACTION

摘要

This volume discusses approaches whereby Neurotron Networks can be used to provide pattern recognition, autonomous decision-making, and action. Our intent is to illustrate the versatility of Neurotron networks, and their potential application to actual problems, and to provide a framework for the mathematical studies of PSV systems. To illustrate pattern recognition, an artificial fovea with jitter analogous to the human eye is described, and we describe how this, together with a Neurotron network, can learn to assign meaning to symbols, including the ability to learn to recognize hand-printed letters. To illustrate autonomous decision-making, it is shown how a Neurotron network can develop its own strategy for playing chess. To illustrate action, it is shown how a Neurotron network can learn to control an arm and hand with visual, tactile, and kinesthetic feedback, and it is shown how a Neurotron network can learn to drive suitable output devices to mimic a simple tune. We also consider locomotion over rough terrain, and the unification of pattern recognition, autonomous decision-making, and action in a single integrated piece of equipment such as an artificial animal or, ultimately, a robot.