Winner Trace Marking in Self-Organizing Neural Network for Classification

The classification for similar features classes is quite difficult task in many existing pattern-recognition systems. When the amount of samples is insufficient, neural networking training is hard. The dimension reduction, classification, clustering etc serial steps in recognition process takes such much time that the practical recognizing application is ease to meet the real time requirement. The new method is looking forward to. This paper presents a fast, simple and robust classifier, in which the winner has been traced and marked during entire training. We named it as Winner Trace Marking (WTM). The basic structure is based on self organizing feather map (SOFM), but the training and recognizing rules are changed and optimized. By WTM, a significant improvement is reached about above problems. The accuracy is highly increased with less time consumption. The experiment classifying strip surface defects by WTM are presented. The results are satisfactory.