Complexity Analysis For The First Class B-Spline Weight Function Neural Networks And Its Application

The standard measure of good or bad for an algorithm is the complexity, the theory analysis ind research of algorithm complexity has the extremely vital significance for the promotion and application of the algorithm. The main purpose of study the first class B-spline weight function neural network is used for demonstrating the algorithm is more excellent than the traditional neural network algorithm. That’s provide theoretical reference for the first category B-spline weight function neural network algorithm in the application of engineering.The traditional neural network has disadvantages such as misconvergence or slow convergence speed, and easy to fall into local minimum etc. In monograph "New Theory and Methods of Neural Network" put forward a new type of structure of the neural network to solve these problems. The research of this paper is based on the spline weight function on the related concepts of neural network, combining with B-spline curve related to the nature of the orrelation analysis method and the complexity of algorithm, the first kind of B-spline weight function is deduced formula of the complexity of the neural network algorithm.We found the algorithm complexity of the first B-spline weight function is not only determined by the dimension of the input layer and output layer, and associated with the number of splines. When the dimension of input layer and output layer is invariable, increase the interpolation spline, the execution time of the algorithm is exponentially. When the dimension of output layer and the spline interpolation is invariable, the execution time of the algorithm is linear. When the dimension of input layer and the spline interpolation is invariable, the execution time of the algorithm is linear.Theoretical analysis is verified by Matlab simulation and experimental results. From practice to justifying for the first category B-spline weight function neural network, the correctness of the theoretical analysis results, compared with traditional neural network reflects the significant performance advantages.Finally, the first kind of B-spline weight function neural network algorithm is applied to image compression and simulation. The experimental results show that, compared with the traditional neural network, the first kind of B-spline weight function neural network algorithm has better compression ratio, faster compression speed and reconstruction image quality has some advantages, such as higher in image compression.