应用交替线性化方法解决支持向量机问题.pdfVIP

第 3O卷第 5期 大 学 数 学 Vo1．30，No．5 2014年 10月 COLLEGE MATHEMATICS 0ct．2Ol4 AnAlternatingLinearizationM ethodforSparse SupportVectorMachine L ， BAjXi。 (1．SchoolofScience，QiqiharUniversity，Qiqihar，Heilongjiang161006 China： ， 2·P·C．RossinCollegeofEngineeringandAppliedScience，LehighUniversity， Beth1ehem ，PA 18015，USA) Abstract：Weproposeanew numericaloptimizationapproach-- alternatinglinearizationmethod (ALM ) forl～ 1supportvectormachinemode1，based on theclassicangmented Lagrangian framework ． Num erical resultstosolvesomedigitrecognitionproblemsprovetheeffectivenessofourmethod ． Key words： numerical optimization； convex optimization； support rector machine； ahernating linearizationmethod；augmentedLagrangianmethod CLC Number：0224 DocumentCode：A ArticleID：1672．1454(2014)05—0001．07 1 IntrOductjOn SupportVectorM achines (SVM )havebeen proved to be an effective classification toolin machinelearning．Itisoriginallyfrom aconceptinstatisticsandcomputerscienceforasetofrelated supervised learning methods thatanalyzedata and recognizepatterns，used forclassification and regressionanalysis．TheconceptofSVM isoneofthesubjectsinmachinelearning，whichisabout learningstructurefrom data．Ithasempiricallygoodperformanceaswellassuccessfu1applicationsin manyfields，suchasbio—informatics，text，imagerecogniti