Study And Application Of Neural Network With Spline Weight Fuction

Artificial neural network is a network topology structure composed of many simple processing inter-comnected units, whose characters are highly adaptive, self-learning ability and non-linear global role. We can achieve any complex function through processing simple units in above network topology structure. First, this paper introduces the basic theory of the neural network and cubic spline weight functions. Second, it introduces the application of the script recognition with neural network and spline weight functions neural network respectively, and the main research can be listed as follow:Firstly, we establish a score prediction model of a university by the spline Weight functions neural network whose structure only consist of the input layer and the output layer. We adopt subsection of cubic sample weight functions as a network of weight function. First we choose five group datas of the experiments as a training network sample, and apply the piecewise cubic spline interpolation weight functions to each sample interval, which can determine the corresponding interval network interpolation coefficient the corresponding interval network. In addition we choose twenty extra group datas of the experiments as our simulative datas, because we have gained the interpolation coefficient which is corresponded with front interval network. Finally we plus the output of each input data as the total output, which can be used as the predicted value of our network training, compared with the calculated real value, and give the error between the two value. We can get a good experimental effect.Besides, according to neural network with the stability and high distinction, the characteristics of character recognition technology of CMAC neural network can be put forward to, and the handwritten character recognition system interface could be made, which the theoretical practice is based on. Simulative experiments show that recognition rate and text features are greatly improved, and this method can be applied to distinct many similar symbols. Based on this discussion, Character recognition technology is improved by spline weight functions neural network, making the recognition result faster and better.Lastly, we summarize the research work of the full text and give the prospect of the future work.