A Study On Improved SASONN

Clustering and Classification is a hot research field and a tough job in statistics, machine learning, and pattern recognition. Meanwhile, Neural Network is an efficient way to solve these problems because it has a special character of perception that the characteristic which it interested in can be grasped. I have made my effort to analyze the Structure-Adapting Self-Organizing Neural Network (SASONN) in the prospect of theory, algorithm and computer testing. In the thesis I firstly introduce three important model in the class of Self-Organizing Neural Network: competitive learning, Adaptive Resonance Theory (ART) and Self-Organizing Mapping (SOM). Then I introduced the Clustering and Classification theory and SASONN model which was proposed by Yin Wu in the following chapter. Finally I summarized these current theory and point out the advantage and disadvantage it. The algorithm I proposed has a lot of parameters and they can satisfactorily integrated with computer technology therefore it don't need a lot of computation intension. Testing result shows that this algorithm works well in Clustering and Classification and has a good Generalization Capability. There is no doubt that the Improved SASONN model proposed in this thesis can be widely used in many fields such as person identification, facial expression synthesis and analysis, multi-modal human computer interaction, etc.