perceptron_training.m

来自「基于感知器算法的线性分类程序」· M 代码 · 共 38 行

function W=Training(Sample1,Sample2);
[N1,l]=size(Sample1);
[N2,l]=size(Sample2);
W=zeros(l+1,1);
S1=ones(l+1,N1);
S1(1:2,:)=Sample1';
S2=ones(l+1,N2);
S2(1:2,:)=Sample2';
k=0;
r=1.5
J=1;
while (J>0&k<100)
    k=k+1
    dJ=zeros(l+1,1);
    Nerror=0;
    for i=1:N1
        if W'*S1(:,i)<=0
            dJ=dJ-S1(:,i);
            Nerror=Nerror+1;
        end
    end
    for i=1:N2
        if W'*S2(:,i)>=0
            dJ=dJ+S2(:,i);
            Nerror=Nerror+1;
        end
    end
    if Nerror>0
        W=W-r*dJ
    else
        J=0;
    end
%     x=[-10:10];
%     y=(-W(1)*x-W(3))/W(2);
%     plot(x,y);
    Nerror
    J
end