The Research On Geometrical Learning Algorithm Based On Biomimetic Pattern Recognition

Since biomimetic pattern recognition was proposed by Wang Shoujue,it has been widely applied the in many fields, such as pattern recognition,face recognition, iris recognition and adaptive controller. In this thesis, wefirstly discuss a kind of new pattern recognition methods -geometrymethod, and give a geometry analysis method for the analysis of artificialneural network. For example, ART neural network, RBF neural network,DBF neural network, and two-weighted neural network etc. Secondly, wediscuss the importance of geometrical learning algorithm through PCAanalysis method and prove its validity by experience.Biomimetic Pattern Recognition is a new model of PatternRecognition based on "matter cognition" instead of "matter classification".The main principle is how to cover the same samples' distribution area.We firstly definite hypersausages neuron based on Biomimetic patternrecognition, and study its structure and geometry algorithms. Third, we propose min-sphere covering method based on Biomimetic patternrecognition and apply for face recognition. Experiments prove its validity.