| Optimization Algorithm is such an technology,based on mathmetics and used for solving the optimized solution for engineering control .It has developed to be an important branch of Science,paid on more and more attention,used extensitively in System Enginerring,Artificial Inteligence,Pattern Recogintion,Produce Dispatching etc.On the other hand,Optimization Technology has found its extensive usage in embeded system based terminal equipment,such as Hand-Writing Recoginiton,Voice Recognition.But these applications just implement one part of Optimization Technology on the embeded system,which is called generalization.This paper researchs the implementation of the training method of optimization algorithm on the embeded system.Two Optimization Algorithms and their respective implementation are present in this paper.One is Neural Network;Neural Network has lots of good features such as parallel processing,fault-torlerance,self-organization,self-adapting etc.This method transfers the actual problem into the evolvement of Neural Network.In this method,the optimization solution is corresponded to the stable state of the Neural Network.Another Optimization Alogrithm is Support Vector Machine;it has presented good performance in solving actual problems.This method is used extensive in the Pattern Recognition,Approximation of Function etc.This paper chooses typical implementations of the two Optimization Algorithm,BP algorithm for Neural Network, SVM algorithm for Support Vector Machine.Optimization Algorithm is core computation of Inteligence Algorithm .So the Research on the Optimization Algorithm's embeded usage can help more Inteligence Algorithm to be used on embeded system.First,new exp function is used to improve the performance of the Inteligence Algorithms.Then,some key conditions are analyzed and modified properly .As a result,the performance of the improved algorithm is improved exponentially.For the reason that Optimization Algorithm is core computation of Inteligence Algorithm,we find the similar transformation on the SVM algorithm.In addition,compared with this weak point,the gap of the CPU frequency is not so important.I believe that if we want to make the Optimization Algorithm have good performance on the embeded system,hardware solution is necessary. |