rbf.m

实现了RBF网络

%径向基网络 采用均值聚类方法，隐含层输出函数采用高斯函数
a=load('a.txt');
t=load('t.txt')
p=a';
t1=t';
tic
samnum=528;
testsamnum=528;
indim=10;
outdim=4;
clusternum=2;%13个聚类中心
overdap=1.5;%重叠系数
centers=p(:,1:clusternum);%初始中心
numberinclusters=zeros(clusternum,1);%各类中的样本数，初始化为零
indexinclusters=zeros(clusternum,samnum);%各类所含样本的索引号
while 1
numberinclusters=zeros(clusternum,1);%各类中的样本数，初始化为零
indexinclusters=zeros(clusternum,samnum);%各类所含样本的索引号
%按最小距离原则对样本进行分类
for i=1:samnum
alldistance=dist(centers',p(:,i));
[mindist,pos]=min(alldistance);
numberinclusters(pos)=numberinclusters(pos)+1;%最小距离数的个数加1
indexinclusters(pos,numberinclusters(pos))=i;%索引号
end
oldcenters=centers;                     %保存旧的聚类中心
for i=1:clusternum                      %重新计算各类的聚类中心
index=indexinclusters(i,1:numberinclusters(i));
centers(:,i)=mean(p(:,index)')';
end
equalnum=sum(sum(centers==oldcenters));%判断聚类中心是否一致
if equalnum==indim*clusternum;
break,
end
end
%计算各隐节点的扩展常数
alldistances=dist(centers',centers);%计算隐节点数据中心的距离
maximum=max(max(alldistances));
for i=1:clusternum
alldistances(i,i)=maximum+1;
end
spreads=overdap*min(alldistances)';
distance=dist(centers',p);
spreadsmat=repmat(spreads,1,samnum);
hiddenunitout=radbas(distance./spreadsmat);
hiddenunitoutex=[hiddenunitout' ones(samnum,1)]';
w2ex=t1*pinv(hiddenunitoutex);
w2=w2ex(:,1:clusternum);
b2=w2ex(:,clusternum+1);
testdistance=dist(centers',p);
testspreadsmat=repmat(spreads,1,testsamnum);
testhiddenunitout=radbas(testdistance./testspreadsmat);
s=w2*testhiddenunitout+repmat(b2,1,testsamnum)
w2
b2
s                %最后的输出
for i=1:4
  for  j=1:528
 if  s(i,j)>=0.52
 s(i,j)=0.99;
 else  s(i,j)=0.01;
 end 
 end
 end
count(1,1:11)=0;
for j=1:528
if s(:,j)==t1(:,j)
 if s(:,j)==[0.01;0.01;0.01;0.99]
  count(1)=count(1)+1;
 elseif  s(:,j)==[0.01;0.01;0.99;0.01]
  count(2)=count(2)+1;
 elseif   s(:,j)==[0.01;0.01;0.99;0.99]
  count(3)=count(3)+1;
 elseif  s(:,j)==[0.01;0.99;0.01;0.01]
  count(4)=count(4)+1;
 elseif   s(:,j)==[0.01;0.99;0.01;0.99]
  count(5)=count(5)+1;
 elseif  s(:,j)==[0.01;0.99;0.99;0.01]
  count(6)=count(6)+1;
 elseif   s(:,j)==[0.01;0.99;0.99;0.99]
  count(7)=count(7)+1;
 elseif  s(:,j)==[0.99;0.01;0.01;0.01]
  count(8)=count(8)+1;
 elseif   s(:,j)==[0.99;0.01;0.01;0.99]
  count(9)=count(9)+1;
 elseif  s(:,j)==[0.99;0.01;0.99;0.01]
  count(10)=count(10)+1;
 elseif   s(:,j)==[0.99;0.01;0.99;0.99]
  count(11)=count(11)+1;
  else  end  
end
   end 
COUNT(1,1:11)=0;
for j=1:528
if s(:,j)==[0.01;0.01;0.01;0.99]
  COUNT(1)=COUNT(1)+1;
 elseif  s(:,j)==[0.01;0.01;0.99;0.01]
  COUNT(2)=COUNT(2)+1;
 elseif   s(:,j)==[0.01;0.01;0.99;0.99]
  COUNT(3)=COUNT(3)+1;
 elseif  s(:,j)==[0.01;0.99;0.01;0.01]
  COUNT(4)=COUNT(4)+1;
 elseif   s(:,j)==[0.01;0.99;0.01;0.99]
  COUNT(5)=COUNT(5)+1;
 elseif  s(:,j)==[0.01;0.99;0.99;0.01]
  COUNT(6)=COUNT(6)+1;
 elseif   s(:,j)==[0.01;0.99;0.99;0.99]
  COUNT(7)=COUNT(7)+1;
 elseif  s(:,j)==[0.99;0.01;0.01;0.01]
  COUNT(8)=COUNT(8)+1;
 elseif   s(:,j)==[0.99;0.01;0.01;0.99]
  COUNT(9)=COUNT(9)+1;
 elseif  s(:,j)==[0.99;0.01;0.99;0.01]
  COUNT(10)=COUNT(10)+1;
 elseif   s(:,j)==[0.99;0.01;0.99;0.99]
  COUNT(11)=COUNT(11)+1;
 
  else  m=0;
   end
 end
CIUNT(1,1:11)=0;
for j=1:528
if t1(:,j)==[0.01;0.01;0.01;0.99]
  CIUNT(1)=CIUNT(1)+1;
 elseif  t1(:,j)==[0.01;0.01;0.99;0.01]
  CIUNT(2)=CIUNT(2)+1;
 elseif   t1(:,j)==[0.01;0.01;0.99;0.99]
  CIUNT(3)=CIUNT(3)+1;
 elseif  t1(:,j)==[0.01;0.99;0.01;0.01]
  CIUNT(4)=CIUNT(4)+1;
 elseif   t1(:,j)==[0.01;0.99;0.01;0.99]
  CIUNT(5)=CIUNT(5)+1;
 elseif  t1(:,j)==[0.01;0.99;0.99;0.01]
  CIUNT(6)=CIUNT(6)+1;
 elseif   t1(:,j)==[0.01;0.99;0.99;0.99]
  CIUNT(7)=CIUNT(7)+1;
 elseif  t1(:,j)==[0.99;0.01;0.01;0.01]
  CIUNT(8)=CIUNT(8)+1;
 elseif   t1(:,j)==[0.99;0.01;0.01;0.99]
  CIUNT(9)=CIUNT(9)+1;
 elseif  t1(:,j)==[0.99;0.01;0.99;0.01]
  CIUNT(10)=CIUNT(10)+1;
 elseif   t1(:,j)==[0.99;0.01;0.99;0.99]
  CIUNT(11)=CIUNT(11)+1;
 
 else      end
end 
for   i=1:11
pc(i)=count(i)/COUNT(i);
pe(i)=count(i)/CIUNT(i);
end
time=toc
pc  
pe 
shijian=cputime