ART Neural Network is based on the Adaptive Resonance Theory. After analyzing the traditional structure of the ART Neural Network, We find that this kind of structure has an inherent property of clustering, but some important information in the input sample is lost, that is, the amplitude of the sample is ignored. So an improved structure and its associated algorithm is presented, and is used successfully to implement a classifier. In the new structure, a new kind of orienting subsystme is added. When sampl...