Greedy Stagewise Algorithm Research Based On SVM

Support Vector Machine (SVM) is a kind of novel machine learning methods Based on Statistical Learning Theory (SLT), which proposed in 1990s by Vapnik, etc. Because of its excellent learning performance and complete theory foundation, SVM has been the hot topic of machine learning and also have successful applications in many fields, such as pattern recognition, data mining, bioinformatics, etc. In comparison to its complete theory foundation, the development of its algorithm has lagged.Especially for its speed and accuracy on training large-scale data set urgent needs to be improved.In this paper, we mainly discuss the training algorithm of SVM on large-scale data set. At first, we introduced the basic concept of SVM theory; and then we systematically described training algorithm of SVM on large-scale data set, and introduce popular SMO algorithm in detail; and then studied SVM training algorithm with the greedy stagewise and reducing methods for SVM to deal with large-scale data set. On the basis of the above, a new greedy stagewise SVM algorithm—YGS-SVMs is presented. The main research in this paper can be classed as follows:(1) Systematically described the basic concept of SVM theory and its main contents.(2) Make the research Systematically of SVM training algorithm to deal with large-scale data set, and make a thorough study to SMO algorithm in.(3)Thorough studied the greedy stagewise SVM training algorithm and reducing methods dealing with large-scale data set. On the basis of the above, we present a new greedy stagewise SVM algorithm, SVMs'greedy stagewise training algorithm based on the large-scale training set rapid reduction, which can be applicable to handle a large-scale data.