EXTRACTION OF FEATURE ATTENTIVE REGIONS IN A LEARNT NEURAL-NETWORK

This paper proposes a new method of extracting feature attentive regio ns in a learnt multi-layer neural network. We define a function which calculates the degree of dependence of an output unit on an input unit . The value of this function can be used to investigate whether a lear nt network detects the feature regions in the training patterns. Three computer simulations are presented: (1) investigation of the basic ch aracteristic of this function; (2) application of our method to a simp le pattern classification task; (3) application of our method to a lar ge scale pattern classification task.