An improved ART2 neural network clustering algorithm
一种改进的ART2网络学习算法

By analyzing ART2 neural network clustering algorithm, an improved ART2 neural network clustering algorithm was proposed. This algorithm applied the sample data to initialize its weights, then designed the formula of modifying weights between feature representation layer and class representation layer based on the principle that the distance between each data in the same class and their clustering center was nearer, clustering result was better. The amplitude information was used by comparing module of input data and clustering center. The analysis illustrates that this algorithm can solve pattern drifting, use amplitude information well, and arise clustering speed.