PATTERN-RECOGNITION IN THE NEOCOGNITRON IS IMPROVED BY NEURONAL ADAPTATION

We demonstrate that equipping the neurons of Fukushima's neocognitron with the phenomenon that a neuron decreases its activity when repeated ly stimulated (adaptation) markedly improves the pattern discriminator y power of the network. By means of adaptation, circuits for extractin g discriminating features develop preferentially. In the original neoc ognitron, in contrast, features shared by different patterns are prefe rentially learned, as connections required for extracting them are mor e frequently reinforced.