The Algorithm Complexity Anylysis And Application Of The Linear Denominator Cubic Rational Spline Weight Function Neural Network

There are many standards to evaluate one algorithm, and complexity is the important one of the standards. A good algorithm can save time cost to solve the problem. The purpose of this paper is to deduce the expression of the algorithm complexity of the linear denominator cubic rational spline weight function neural network by combining the weight function neural network theory and complexity of the algorithm, and then analyze the factors affecting the complexity and make experimental verification. On the basis of theoretical research, the type of neural network algorithm is at last applied to practice.On the theoretical basis of weight function neural networks, this paper combines the character of Hermite interpolation and rational spline interpolation to construct form of linear denominator cubic rational spline weight function. Then we Then according to Peano Kernel Theorem, the definition of the algorithm complexity, knowledge of LU methods, through calculation of the executing times of various types of operation during the algorithm executing process, we get the time complexity of the algorithm. Finally, based on the theoretical analysis, by executing MATLAB simulation experiments, complexity of the linear denominator cubic rational spline weight function neural network algorithm is verified.Through theoretical analysis and experimental verification, we get that time complexity of the linear denominator cubic rational spline weight function neural network algorithm has the linear relationship respectively with the number of training samples, the network input dimension and the dimension.Through experimental verification,we get that the time complexity of the linear denominator cubic rational spline weight function neural network algorithm has the linear relationship respectively with the number of training samples, the network input dimension and the dimension, and the the algorithm also has low training time complexity and the higher training speed.In this paper, on the basis of theoretical analysis and experimental verification of time complexity of the linear denominator cubic rational spline weight function neural network algorithm, the linear denominator cubic rational spline weight function neural network algorithm is applied to predict the missing sensed data of wireless sensor network. By MATLAB simulation experiments, we see the algorithm has high accuracy in predicting the missing data, the forecasted data has some credibility and reference value.