FactAffinity

 [Cliques,deep]=FactAffinity(MI,vars,sizeconstraint,p,deep)
 FactAffinity:   Finds a set of cliques from a matrix of mutual information
                using affinity propagation
 INPUTS:
 NumbVar: Number of variables
 dim: Number of previous variables each variables depends on 
 OUTPUTS:
 Cliques: Structure of the model in a list of cliques that defines the (chain shaped)  junction tree. 
          Each row of Cliques is a clique. The first value is the number of overlapping variables. 
          The second, is the number of new variables.
          Then, overlapping variables are listed and  finally new variables are listed.
 Last version 8/26/2008. Roberto Santana and Siddarta Shakya (roberto.santana@ehu.es)

0001 function [Cliques,deep]=FactAffinity(MI,vars,sizeconstraint,p,deep)
0002 % [Cliques,deep]=FactAffinity(MI,vars,sizeconstraint,p,deep)
0003 % FactAffinity:   Finds a set of cliques from a matrix of mutual information
0004 %                using affinity propagation
0005 % INPUTS:
0006 % NumbVar: Number of variables
0007 % dim: Number of previous variables each variables depends on
0008 % OUTPUTS:
0009 % Cliques: Structure of the model in a list of cliques that defines the (chain shaped)  junction tree.
0010 %          Each row of Cliques is a clique. The first value is the number of overlapping variables.
0011 %          The second, is the number of new variables.
0012 %          Then, overlapping variables are listed and  finally new variables are listed.
0013 % Last version 8/26/2008. Roberto Santana and Siddarta Shakya (roberto.santana@ehu.es)
0014 
0015 
0016 if deep>=10
0017    MI = MI + rand(size(MI)); 
0018 end
0019 deep = deep + 1;
0020 n = size(MI,1);
0021  vars
0022 [idx,netsim,dpsim,expref]=apcluster(MI,p); %Find the clusters of the MI matrix
0023 Cliques = [];
0024 nclusters = length(unique(idx));
0025 exemplars = unique(idx);
0026 for i=1:nclusters
0027  clusters{i} = find(idx==exemplars(i));
0028  clustsize = size(clusters{i},1);
0029  if(clustsize>sizeconstraint)
0030     MI1 = MI(:,clusters{i});
0031     MI1 = MI1(clusters{i},:);
0032     p=  (max(MI1(:))-min(MI1(:)))*rand();  % Set preference to median similarity
0033     olddeep = deep;
0034     [Cliques1,deep]=FactAffinity(MI1,vars(clusters{i}'),sizeconstraint,p,deep);
0035     deep = olddeep;
0036    Cliques = [Cliques;Cliques1];
0037  else
0038    %vars(clusters{i}')'
0039    Cliq = [0,clustsize,vars(clusters{i}'),zeros(1,sizeconstraint+2-clustsize)];
0040    Cliques = [Cliques;Cliq];
0041  end
0042 end 
0043 
0044 
0045 
0046 
0047 
0048 
0049 
0050 
0051 
0052