Learning to control dynamic systems via associative reinforcement learning

and the lack of explicit instructional information about how to perform a given control task. Under these circumstances, techniques developed by arti cial intelligence researchers for \learning from examples," including the \supervised learning" techniques studied by neural network researchers, are not directly applicable because these techniques are based on the availability of training information (the \examples") in the form of situation-action training pairs. A useful alternative in such circumstances is a learning technique that can discover