mainsvc.asv

这个是支持向量聚类机用matlab编译的主程序和部分子程序

%=====================================================================
% 
%  Final Project in Machine Learning
%  ----------------------------------
%
%  Submitted By:
%   
%      Ofer Pasternak
%      Elhanan Borshtain 
%      Orit Kliper
%
%  -------------------------
%  Support Vector Clustring 
%  -------------------------
%	A non-parametric clustring algorithm based on support vector 
%	approach.
%
%	Reference: 
%	A.Ben-Hur,D.Horn,H.T.Siegelmann and V.Vapnik.
%	Support Vector Clustring.
%	Journal of Machine Learning Research 2(2001)125-137
%
%   Parameters:
%           
%           example_set - integer number coding the wanted Sample set.
%           C           - Defines the fraction of points which are allowed
%                         to become outliers.
%                         (p = 1/CN where N is the the total sampels number).
%           qvec        - vector of q values for multiple runs over the chosen sample set.
%                         each q value is a different width of the gaussian kernel.
%           filename    - a name for a file to hold the results.
%	
%=====================================================================
function [] = MainSVC(example_set,C,qvec,filename);
clc;
clear;
time1=cputime;
% start clock
tic

count_q = 0;

% 自己设置的参数 2006.1.7
example_set=3;
C=500;
q=10;

    
% Samples from a given example set: 
% seperate the Sample set and the classification vector
% also reads the samples number, and dimention
[Samples, N, classifications] = ReadData(example_set);
if Samples == 0
    fprintf('invalid example set\n');
    return;  
end


% Each Sample set is analyzed using different q value, set by qvec.
%for q=qvec
    
    % Preforms Support Vector Clustering  
    [SV,BSV,beta,quad,R] = SVC(Samples,C,q);
    if(R == -1) return; end


    % Finds the clusters assignments
    [clusters_assignments,maj_class, mis_class,nof_samples_per_class_per_cluster] =...
        SeperateClusters(Samples,beta,quad,R,q,classifications);
    
    %Maps the Sphere back to the data space 
    [grids, grids_sizes] = MappingAnalysis(Samples,R,beta,quad,q);

    % Shows graphic Results:
    ShowResults(Samples,SV,BSV,R,beta,quad,q,clusters_assignments, grids, grids_sizes,classifications,nof_samples_per_class_per_cluster);

    count_q = count_q+1;
    mis_per_q(count_q) = 100*sum(mis_class)/N;
    cluster_per_q(count_q) = max(clusters_assignments);
    
    %save (strcat(filename,'q',num2str(q)));
    MisPercentQ=100*sum(mis_class)/N;
    ClusterPerQ = max(clusters_assignments);
    runtime=cputime-time1;
    save SVCresult MisPercentQ  ClusterPerQ clusters_assignments runtime
%end

% shows Total results varying over q values.
%figure;
%plot(qvec,mis_per_q);
%title('Misclassifications over Ranging Values of q');
%xlabel('q');
%ylabel('Misclassifications (%)')

%figure
%plot(qvec,cluster_per_q);
%title('Change of the Number of Clusters over Ranging Values of q');
%xlabel('q');
%ylabel('Number of Clusters');