A selective attention template matching neural network

A kind of similarity measure with a selective attentional property is proposed. The neural network (selective attentional template matching network, SATMN) model built with the proposed measure can be applied to template matching and classification. The model is composed of matching subnets and competing subnets, in which comparison and classification are carried out respectively. The similarity measure is available to both discrete and analog patterns. The experimental results show that it takes advantage of several widely used similarity measures in describing the variance between patterns. The neural network implementation and experimental results are given.