mk_nbrs_of_pdag_del.m.svn-base

来自「bayesian network structrue learning mat」· SVN-BASE 代码 · 共 65 行

function [PDAGs, nodes] = mk_nbrs_of_pdag_del(cpdag)
% MK_NBRS_OF_PDAG_ADD Make the inferior inclusion boundary of CPDAG.
% [PDAGs, nodes] = mk_nbrs_of_pdag_del(CPDAG)
%
% PDAGs{i} is the i'th neighbor of CPDAG0 generated by DELETE(X,Y,H)  with
% nodes{i,1:2}=[X Y]
% nodes{i,3}=H
%
% See D.M. Chickering 2002 : "Optimal Structure Identification with Greedy Search".
%
% [e-mail]
% 25 july 2003

compteur=0 ;
N=length(cpdag);
bnet_tmp=mk_bnet(pdag_to_dag(cpdag),2*ones(N,1));
engine_tmp=struct(jtree_inf_engine(bnet_tmp));
cliques=engine_tmp.cliques;
nbcliques=length(cliques);
clear bnet_tmp engine_tmp;

verbose=0;

if verbose
    fprintf('---- Les cliques (maximales) du DAG\n');
    for i=1:nbcliques,
        disp(cliques{i});
    end
    fprintf('---- Variables \n');
end


% find in the PDAG all the X Y connected
[LX LY]=find(cpdag==1);
nlinks=length(LX);

for i=1:nlinks
    X=LX(i);
    Y=LY(i);
    % Neighbors of Y
    NY = myintersect(find(cpdag(:,Y)), find(cpdag(Y,:)));
    % Adjacents of X
    AX =  myunion(find(cpdag(:,X)), find(cpdag(X,:)));
    % Neighbors of Y adjacent to X    
    NAYX = myintersect(NY,AX);
    
    % this function recursively "walks" (dfs) in the graph representation of NA powerset
    if verbose
        X, Y, NY, AX, NAYX
        fprintf('---- NA Powerset\n\n');
    end
    
    liste=NAYX;    
    if ~isempty(liste)
        premier=liste(1);
        dernier=liste(end);
    end;     
    
    current_set=[];
    fini=0;
    evite2 = 0 ;
    
    while ~fini
        if verbose
            % sert uniquement