The Researcher On Theory Of Structural Approximation Of Double Weights Neural Network And Its Application

The double weights neural network, which was proposed by academician Wang shoujue, is a new neural network based on the (connective) network theory. The double weights neural network own two weights, the feed forward neural network which constructed by the double weights neural network, realizes a superior capability function approach network. Presently, the double weights neural network is applied widely to face detection, Speech recognition, Proteins structure prediction, and so on.Firstly, in this paper we reviewed the phylogeny and actuality of neural network, and summarized basal conceptions and theories of it, we also described the mathematical model, characteristics and the algorithm of the double weights neural network. Then, the function approaches theory of neural network was analyzed and summarized.