ACTION‐ORIENTED LEARNING NETWORKS

This paper describes the use of the MINERVA adaptive computing system in an investigation of feedback in learning networks. The feedback is such as to provide, at the input of the network, information regarding the action which the network is taking as a result of a pattern which is input on a different set of terminals. It will be shown that this type of network is very sensitive to small differences between incoming patterns and can provide the basis of a model of animal behaviour such as may be found in frogs when snapping at an insect.