Research And Application Of Classification Neural Network Based On Extension

The researchful object of the extenics is the existing contradiction problem in the realistic word, whose researchful orientation is rules and methods of contradiction problem. Extenics try to formalize the process when people settle contradiction problem and establish the extension model in the same time. Based on these knowledge, some new method of design and technique of engineering are produced. Neural network is a complex system that is made up of large numbers of mutual connecting single cell. It can simulate the structure and function of cerebra and also has characters of parallel computation , nonlinear process , self study, and so on. One neural network can have much different kind of information. Every connection weight of neural network has part of the information.When neural network is proposed, it has been applied in many fields such as information , automation, engineering, medicine, economy. Neural network has many merits, but the number of layer and nodes in every layer does not have ideal theory. The structure of neural network is determined by outcome which training data runs in the neural network. The number of algorithm of neural network has more than one hundred, but every algorithm has its limitation.Neural cell and extension element can have different character information, so they have some connection. In this paper, the new definition of neural network is produced, extenics and neural network combine, two extension classification neural network structures and corresponding algorithm are produced. The choice of the number of input node and output node is also discussed. So we can design two different extension classification neural network (ECNN) models. We can acquire better result with the procedure of matlab 6.5. Its operational speed is very rapid, All these prove thefeasibility of two models.