Algorithms Comparsion For Solving LS-TSVR And Minimal Error Margin SVR

This paper discussed the advantages of fast algorithm in different least squares twin support vector regression machines(LS-TSVR),and the application of large margin theory on support vector regression machine(SVR).We studied the fast algorithm for solving least squares twin bounded support vector regression model(LS-TBSVR);And then,we applied this algorithm to solving the least squares multiple output regression problems.In order to verify the effectiveness of the fast algorithm,we compared with the classical algorithm made a series of experiments.This paper mainly divided into four chapters.The first chapter introduced the SVR models that used in this paper,and analyzed the superiority and inferiority of each model.The second chapter applied the large margin theory to support vector regression machine,defined the interval error,proposed the minimal error margin SVR and minimal error margin TSVR,and gave the learning algorithms,respectively.We analyzed their theoretical advantages for solving practical problems,simultaneouslyThe third chapter comparsed the two algorithms for solving LS-TBSVR,and gave the corresponding algorithms steps.Meanwhile,We verified the effectiveness of fast learning algorithms by experiments,and proved the advantages of fast algorithm in computing speed and precision.The fourth chapter studied the fast algorithm for solving multioutput twin least squares regularized support vector regression model(M-TLS-RSVR)and multioutput least squares twin bounded support vector regression model(M-LS-TBSVR),proposed two novel fast learning algorithms,and gave the specific algorithm steps,respectively.We discussed the advantages and disadvantages of them in solving the multi-uotput regress problem by experiments.We also proved that the fast algorithm can be used to solve multi-output problems.