matlab实现简单c均值算法.pdfVIP

Matlab 实现C 均值： clear;clc; %类数，c2 c=3;pointnum=50; pointheart=[30,30;80,40;50,70]; %高斯分布产生二维随机数 point =[mvnrnd (pointheart (1,:),[30,0;0,30],pointnum);... mvnrnd (pointheart (2,:),[30,0;0,30],pointnum);... mvnrnd (pointheart (3,:),[30,0;0,30],pointnum)]; %增广化 [x,y]=size(point); point=[point,ones (x,1)]; %ρ(用p 表示) p=1; %w1;w2;w3 w=ones (c,3); %开始感知器算法 flag=1; while (flag==1) flag=0; fori=1:c for ii = point ((i-1)*pointnum+1:i*pointnum ,:) d=w*ii; flag2=0; for iii=1:c if (iii==i) continue; end if (d (iii)=d (i)) flag=1; flag2=1; w (iii,:)=w (iii,:)-p*ii; end end if (flag2==1) w (i,:)=w (i,:)+p*ii; end end end end w %求边界,d(i)=d(j) wb=zeros(c,3); xmax=max (point (:,1));xmin=min (point (:,1)); ymin=min (point (:,2)); %求各边界的公共交点 atmp=[]; for xi=1:c-1 foryi=xi+1:c atmp=[atmp;w (xi,:)-w (yi,:)]; end end r=rank(atmp); a=null (atmp,r);%解向量 a 有可能为空 %画边界，仅c=3 情况,更多维考虑atmp 顺序问题 hold on; for xi=1:c if (norm ([xmin-20,-((atmp (xi,1)*(xmin-20)+atmp (xi,3))/atmp (xi,2))]-poi ntheart (c-xi+1,:))=... norm ([xmax+20,-((atmp (xi,1)*(xmax+20)+atmp (xi,3))/atmp (xi,2))]-pointh eart (c-xi+1,:))) xx=(xmin-20):a (1,1); else xx=a (1,1):(xmax+20); end y=-((atmp (xi,1)*xx+atmp (xi,3))/atmp (xi,2)); plot (xx,y,k); end %画点 plot (point (1:pointnum ,1),point (1:pointnum ,2),+r,... point (pointnum+1:2*pointnum ,1),point (pointnum+1:2*pointnum ,2),+b,.. . point (2*pointnum+1:3*pointnum ,1),point (2*pointnum+1:3*pointnum ,2),+g ); %测试,10个点测试 testpoints=rand(10,2)*100; [x,y]=size(testpoints); testpoints=[testpoints,ones (x,1)]; colors=rand(c,3); fori=1:x [maxn ,postion]=max (w*testpoints (i,:)); plot (testpoints (i,1),testpoints (i,2),*,Color,colors (postion ,:)); end hold off; 效果如下，红蓝绿较为密集的点是测试用的点，而比较稀疏的点是用来训练的点。