step2.asv

模式识别中的isodata算法

% step2.m
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% All important data in (cell) array s.
% The first row are s(1,:)centers,then s(2,:)samples,s(3,:)in-class distance!!
% sample classification based on minimum distance princple. 
% So it include Step 4,5,6,8
s=cell(5,c);
for i=1:c
    s{1,i}=Z(i,:);
end

%分类

for i=1:c % 清空s{2.i}
   s{2,i}=[];
end % 分类
for i=1:length(sample)
    [temp minposition]=mindistance(Z,sample(i,:));
    clear temp
    s{2,minposition}(end+1,:)=sample(i,:);
end

for i=1:c %修改聚心 
    if size(s{2,i},1)~=0
        s{1,i}=[mean(s{2,i}(:,1)) mean(s{2,i}(:,2))];
        Z(i,:)=s{1,i};
    end
end

% 统计每类样本数
for i=1:c
   s{5,i}=size(s{2,i},1);
end

%计算类内距离
for i=1:c
    if size(s{2,i},1)~=0
       s{3,i}= inclassdistance(s{1,i},s{2,i});
    end
end

% 计算总平均距离
d_avg=0;
for i=1:c
    d_avg=d_avg+sum(s{5,i}*s{3,i})/size(sample,1);
end

% Step 8
for k=1:c   
   s{4,k}=std(s{2,k},1,1);
   % the second 1 controls std use the form of standard deviation, the
   % third 1 means the 1st demension
   % see more details in the fucntion 'std'  
end

% names
s{1,end+1}='center';
s{2,end}='sample';
s{3,end}='in-class distance';
s{4,end}='standard deviation';
s{5,end}='number of samples';
%%%Author: Feng Shuo%%%%%%%%