一种改进的模糊多类支持向量机算法.pdf

. 2 0 11. 19 ( 4) 908 Computer Measurement Control : 2011) : T P391 : A 李广莉, 崔广顺 ( , 067000) : ; , ; , , , , , , , , : ; ; ; A Improved Algorithms of Fuzzy Support Vector Machines Li Guangli, Cui Guangshun ( D partm nt of Comput r and Information Engin ring , Ch ngD P trol um Coll g , Ch ngD 067000, China) Abstract: Support V ctor Machin s ( SVM ) ar a n w - g n ration machin l arning t chn iqu bas d on th statistical l arning th ory. Th y can solv small- sam pl l arning probl m s b tt r by u sing Stru ctural Risk Minimization in plac of Exp ri ntial Risk Minimization. This pap r bas d on statistic l arning th ory ( SLT ) and support v ctor machin ( SVM) . Aim d at th d f ct in th m thod of fuzzy support v ctor machin w h r th m mb rship function w as con struct d by m ans of distanc r lation b tw n th sampl and clust r c nt r, xt nd to th Multiclass Su pport V ctor Machin s m thod, and w ith th fuzzy m m b rship of data sampl s of a giv n class, to improv classification p rforman c w ith high g n ralization capability. W us d th improv d S qu ntial Min imal Optimization to solv fuzzy multicat gory support v ctor machin . T h xp rim ntal r sults show that th computational load b r du c d gr atly an d w ith high g n ralization capability. Key ords: support v ctor m achin , statistical l arning th ry, multiclassification m thod, fu zzy m m b rship of data sampl s 0 ; ( 2) , , ( , , ) , ,