支持向量機(jī)程序

1、該工具箱包括了二種分類二種回歸以及一種一類支持向量機(jī)算法(1)Main_SVC_C.mC_SVC二類分類算法(2)Main_SVC_Nu.mNu_SVC二類分類算法(3)Main_SVM_One_Class.mOneClass支持向量機(jī)(4)Main_SVR_Epsilon.mEpsilon_SVR回歸算法(5)Main_SVR_Nu.mNu_SVR回歸算法%%3使用(1)目錄下以Main_開頭的文件即是主程序文件直接按快捷鍵F5運(yùn)行即

2、可(2)工具箱中所有程序均在Matlab6.5環(huán)境中調(diào)試通過，不能保證在Matlab其它版本正確運(yùn)行%%%%SupptVectMachineMatlabToolbox1.0CSupptVectClassification%Platfm:Matlab6.5Matlab7.0%Copyright:LUZhenboNavyEngineeringUniversityWuHanHuBeiP.R.China430033%Email:luzhenbo

3、@.cn%Homepage::luzhenbo..cn%Reference:ChihChungChangChihJenLin.“LIBSVM:aLibraryfSupptVectMachines“%%Solvethequadraticprogrammingproblem“quadprog.m“clcclearcloseall%%%定義核函數(shù)及相關(guān)參數(shù)C=200%拉格朗日乘子上界ker=struct(typelinear)%ker=str

4、uct(typeploydegree3offset1)%ker=struct(typegausswidth1)%ker=struct(typetanhgamma1offset0)%ker核參數(shù)(結(jié)構(gòu)體變量)a=svm.aepsilon=1e8%如果小于此值則認(rèn)為是0i_sv=find(aepsilon)%支持向量下標(biāo)plot(X(i_sv1)X(i_sv2)ro)%%%測試輸出[x1x2]=meshgrid(2:0.05:72:0.05

5、:7)[rowscols]=size(x1)nt=rowscols%測試樣本數(shù)Xt=[reshape(x1nt1)reshape(x2nt1)]ticYd=C_SVC_Sim(svmXt)%測試輸出t_sim=tocYd=reshape(Ydrowscols)contour(x1x2Yd[00]m)%分類面holdofffunction[K]=CalcKernel(kerxy)%Calculatekernelfunction.%%x:輸

6、入樣本n1d的矩陣n1為樣本個(gè)數(shù)d為樣本維數(shù)%y:輸入樣本n2d的矩陣n2為樣本個(gè)數(shù)d為樣本維數(shù)%%ker核參數(shù)(結(jié)構(gòu)體變量)%thefollowingfields:%typelinear:k(xy)=xy%poly:k(xy)=(xyc)^d%gauss:k(xy)=exp(0.5(nm(xy)s)^2)%tanh:k(xy)=tanh(gxyc)%degreeDegreedofpolynomialkernel(positivesca

7、lar).%offsetOffsetcofpolynomialtanhkernel(scalarnegativeftanh).%widthWidthsofGausskernel(positivescalar).%gammaSlopegofthetanhkernel(positivescalar).%%ker=struct(typelinear)%ker=struct(typeploydegreedoffsetc)%ker=struct(