Complexiy Of The Algorithm Study And Application Of The Second Category Of B-spline Weight Function Neural Network

The complexity of the algorithm is a measure of the difficulty of the scale of the algorithm,analysis of the complexity of the algorithm is an important issue. The most common evaluationcriteria is the time and space spend by algorithm. For well-implemented algorithm can often bringhigher operating and storage efficiency. So study the second category of B-spline weight functionneural network algorithm complexity is to prove that it is a very good algorithm.Spline weight function neural network is a new structure of the artificial neural network, whichfor the first time proposed in “The new neural networks theoryand method”. This study researchedthe second category of B-spline weight function neural network algorithm complexity based on thespline weight function neural network, combined with the characteristics of weight function, neuralnetwork, attributes of B-spline function and related analytical methods, then deduced the secondcategory of B-spline weight function neural complexity formula of the B-spline.The formula shows that the second category of B-spline weight function neural networkalgorithm complexity have a relationship with the number of times of the B-spline function，inputand output dimension and the number of training sample points. When the high number of times ofinterpolation B-spline basis function, the time complexity of the algorithm grow exponentially.Otherwise, the time complexity of the algorithm grow linear.The contrast experiments with traditional neural network testified that the second category ofB-spline weight function function neural network has better network performance and verified theformula is corrected.Based on the analysis of the second B-spline weight function neural network algorithmcomplexity, this paper applied the second category of B-spline weight function neural network tothe field of license plate recognition, divided the license plate image into individual characters, thenget the result of every characters, the input of neural network used by the characters features withused K-L transform. The experiments on license plate recognition with the second category ofB-spline weight function neural network classifier is correct, the classifier is very simple with goodalgorithm, the cognition rate is very high.