k-means及isodata算法的matlab实现

最近上模式识别课涉及到聚类算法，主要讲了一下kmeans和isodata，这里是我用matlab实现的两个算法的粗略版，还有很大的改进空间，先把原始版放这上面。

function Kmeans(x,k)
n = size(x,1);
mean = cell(k,1);
for i=1:k
    mean{i} = x(i,:);%initialize 
end
while 1
    class = cell(k,1);%clear before calculate
    for i=1:n
        num = Belong2(x(i,:),mean);
        class{num} = [class{num};x(i,:)];
    end
    %calculate new means
    mean_old = mean;
    for i=1:k
        mean{i} =sum(class{i})./size(class{i},1);
    end
    if isequal(mean_old,mean)
        for j=1:k
            fprintf('第%d类:\n',j)
            disp(class{j});
        end
       break;
    end
end
end
function number = Belong2(x_i,means)
    INF = 10000;
    min = INF;
    kk = size(means,1);
    number = 1;
    for i=1:kk
        if norm(x_i - means{i}) < min
            min = norm(x_i - means{i});
            number = i;
        end
    end
end

isodata

function ISODATA(x,K,theta_N,theta_S,theta_c,L,I)
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
    %%%%%%%%input parameters%%%%%%
% x : data
% K : 预期的聚类中心数
% theta_N : 每一聚类中心中最少的样本数，少于此数就不作为一个独立的聚类