test_fkme.m

K-MEAN:经典K均值算法,适用领域：语音识别

% sample command to run fkme
clear all;
load irises.txt               % Load data from a text file
data = irises;                % data to be clustered

nclass=3;                   % number of class
phi=2;                      % fuzzy exponent >1
maxiter=300;                % maximum iterations
toldif=0.000001;            % convergence criterion
distype=3;                  %   distance type:        1 = euclidean, 2 = diagonal, 3 = mahalanobis
scatter=0.2;                % scatter around initial membership

% run fkme
[U, Ue, centroid, dist, W, alfa, obj] = run_fkme(nclass,data,phi,maxiter,distype,toldif,scatter,ntry)
% output:
%   U           = membership matrix
%   centroid    = centroid              centroid(nclass, ndim)
%   dist        = distance matrix       dist(ndata,nclass)
%   W           = distance norm matrix
%   alfa        = extragrade parameter
%   obj         = objective function

% calculate validity 
[fpi mpe S djdphi]=fvalidity(U,W,centroid,dist,nclass,phi);

% calculate confusion index
ci = confusion(nclass,data,U);

% To test the allocate function 
% to allocate say new data into existing centroid 
[U, Ue, dist, obj] = fkme_all(nclass,data,centroid,W,phi,alfa,distype)