Least Squares And Least Squares Support Vector Machines

The thesis studies the incremental and fixed-size training algorithm of Least Squares Support Vector Machine (LSSVM). LSSVM, which was put forward and based upon Support Vector Machine (SVM) by J. A. K Suykens et. al, is an extended-edition of SVM.The standard algorithm of LSSVM are all offline ones, which couldn't be used online. There are few studies on online LSSVM training algorithms. Based on previous research results, the major contents of this thesis are as following:First, the basic theoretical of the LSSVM algorithm has been gone deep into discussed. Then Least Squares Method, LSSVM and Pruning Spares algorithm of LSSVM have been compared by the simulation results.Finally, we researched on the incremental and the fixed-size training algorithms of LSSVM. For the defect in regression predicting, which LSSVM, an improved method is proposed, which highly performance of LSSVM. Simulation results show the validity of the improved algorithm.