Algorithm Of Svm Based On Iteration In Adaboost

Support Vector Machines (SVM for short) is considered as one of the most potential classification techniques, and it's firstly launched in 1963 by the remarkable research team in AT&Bell Lab. under the leadship of Mr.Vanpik. SVM works as one recognition method based on Theory of Statistical Learning and initialy mainly services the field of pattern recognition. And the sustaining researches and applications led to the repid development and perfection of SVM. it spreads to even function-fitting and other related machine learning issues. From then on, it moved forwards with high speed. Until now, it has been successively used in many fields, such as bioinformatics, text and handwriting recognition.Although SVM has been widely used in many aspects, but unfortunately it's not so well acceptable in classification. While AsaBoost arithmetic has a good reputation in the field of Pattern Recognition, which can be considered as the basis of SVM classification. Therefore, My thesis will convey reaserches both SVM and AdaBoost and combination as follows:1)Analysis and application of AdaBoost Arithmetic Theory convince us of advanteges in Pattern Recognition, especially in Classification. If we make best of Ada Boost and create, the Classification permeability of SVM will be highly improved.2)Based on the merits of both arithmetics, it keeps training these information-riched samples in the iteration process. SVM can helps training samples by its minimizing sequence arithmetic, then the richese information samples will be found, which Adaboost mark with property, and will finalize the best property. Therefore SVM turns out on the basis of iteration process of Adaboost.3)This thesis further shows people a definite new combined arithmetic, Adaboost-SVM after millions of analysis and research, it also provides a structure flow chart, leads us all steps to complish. Each time we can find a weak classifier with even much better classification permeability and indicate rate of accuracy for last classification,. These two factors will then give birth to arithmetic design, that is AdaBoost-SVM. 4)After researches and comparison with SVM, it reports that Adaboost-SVM increase the probability of prediction of SVM and better the permeability of adaline. So this combined arithmetic is well worthy further research and full application.