function svm = svmTrain(svmType,X,Y,ker,p1,p2)% SVM Classification:%   svm = svmTrain('svc_c',x,y,ker,C); %   svm = svmTrain('svc_nu',x,y,ker,nu); %% One-Class SVM:%   svm = svmTrain('svm_one_class',x,[],ker,nu);%% SVM Regression:%   svm = svmTrain('svr_epsilon',x,y,ker,C,e); %   svm = svmTrain('svr_nu',x,y,ker,C,nu); % 輸入參數:% X  訓練樣本,d×n的矩陣,n為樣本個數,d為樣本維數% Y  訓練目標,1×n的矩陣,n為樣本個數,值為+1或-1% ker  核參數(結構體變量)% the following fields:%   type   - linear :  k(x,y) = x'*y%            poly   :  k(x,y) = (x'*y+c)^d%            gauss  :  k(x,y) = exp(-0.5*(norm(x-y)/s)^2)%            tanh   :  k(x,y) = tanh(g*x'*y+c)%   degree - Degree d of polynomial kernel (positive scalar).%   offset - Offset c of polynomial and tanh kernel (scalar, negative for tanh).%   width  - Width s of Gauss kernel (positive scalar).%   gamma  - Slope g of the tanh kernel (positive scalar).% 輸出參數:% svm  支持向量機(結構體變量)% the following fields:%   type - 支持向量機類型  {'svc_c','svc_nu','svm_one_class','svr_epsilon','svr_nu'}%   ker - 核參數%   x - 訓練樣本,d×n的矩陣,n為樣本個數,d為樣本維數%   y - 訓練目標,1×n的矩陣,n為樣本個數,值為+1或-1%   a - 拉格朗日乘子,1×n的矩陣% ------------------------------------------------------------%options = optimset;options.LargeScale = 'off';options.Display = 'off';switch svmType    case 'svc_c', %  svc的c:分類                C = p1;        n = length(Y);        H = (Y'*Y).*kernell(ker,X,X);   %矩陣          f = -ones(n,1);  %向量        A = [];            %矩陣          b = [];        %向量        Aeq = Y;          %矩陣        beq = 0;         %向量        lb = zeros(n,1);   % x的下界        ub = C*ones(n,1);  % x的上界        a0 = zeros(n,1);   %初值                [a,fval,eXitflag,output,lambda]  = quadprog(H,f,A,b,Aeq,beq,lb,ub,a0,options); %fval：返回解a處的目標函數值 quadprog：二次規劃函數  a 拉格朗日乘子                   case 'svc_nu', % nu：υ   分類                nu = p1;               n = length(Y);         H = (Y'*Y).*kernell(ker,X,X);        f = zeros(n,1);        A = -ones(1,n);        b = -nu;        Aeq = Y;        beq = 0;        lb = zeros(n,1);        ub = ones(n,1)/n;        a0 = zeros(n,1);                [a,fval,eXitflag,output,lambda]  = quadprog(H,f,A,b,Aeq,beq,lb,ub,a0,options);                        case 'svm_one_class', % 分類                nu = p1;            n = size(X,2);        H = kernell(ker,X,X);        f = zeros(n,1);        for i = 1:n            f(i,:) = -kernell(ker,X(:,i),X(:,i));        end        A = [];        b = [];        Aeq = ones(1,n);        beq = 1;        lb = zeros(n,1);        ub = ones(n,1)/(nu*n);        a0 = zeros(n,1);        [a,fval,eXitflag,output,lambda]  = quadprog(H,f,A,b,Aeq,beq,lb,ub,a0,options);                                case 'svr_epsilon', % epsilon：ε  回歸                C = p1;        e = p2;                n = length(Y);        Q = kernel(ker,X,X);        H = [Q,-Q;-Q,Q];        f = [e*ones(n,1)-Y';e*ones(n,1)+Y'];          % 符號不一樣,決策函數就不一樣,實際上是一回事!        %f = [e*ones(n,1)+Y';e*ones(n,1)-Y'];        A = [];        b = [];        Aeq = [ones(1,n),-ones(1,n)];        beq = 0;        lb = zeros(2*n,1);                       ub = C*ones(2*n,1);        a0 = zeros(2*n,1);                [a,fval,eXitflag,output,lambda]  = quadprog(H,f,A,b,Aeq,beq,lb,ub,a0,options);                    a = a(1:n)-a(n+1:end);    case 'svr_nu',  %  回歸         C = p1;        nu = p2;        n = length(Y);        Q = kernell(ker,X,X);        H = [Q,-Q;-Q,Q];        f = [-Y';+Y'];          % 符號不一樣,決策函數就不一樣,實際上是一回事!        %f = [+Y';-Y'];        A = [];        b = [];        Aeq = [ones(1,n),-ones(1,n);ones(1,2*n)];        beq = [0;C*n*nu];        lb = zeros(2*n,1);                       ub = C*ones(2*n,1);        a0 = zeros(2*n,1);                [a,fval,eXitflag,output,lambda]  = quadprog(H,f,A,b,Aeq,beq,lb,ub,a0,options);                    a = a(1:n)-a(n+1:end);    otherwise,end% ------------------------------------------------------------%% 輸出 svmsvm.type = svmType;svm.ker = ker;svm.x = X;svm.y = Y;svm.a = a';