Home > Mateda2.0 > sampling > FindMPE.m

FindMPE

PURPOSE ^

[mpe_solution,prob_value] = FindMPE(bnet):

SYNOPSIS ^

function[mpe_solution,prob_value] = FindMPE(bnet)

DESCRIPTION ^

 [mpe_solution,prob_value] =  FindMPE(bnet):   
                           Computes the most probable explanation given the Bayesian network
 INPUTS 
 bnet:                Bayesian network
 OUTPUTS
 mpe_solution: Most probable explanation
 prob_value:  Its corresponding probability value

 Last version 8/26/2008. Roberto Santana (roberto.santana@ehu.es)

CROSS-REFERENCE INFORMATION ^

This function calls: This function is called by:

SOURCE CODE ^

0001  function[mpe_solution,prob_value] =  FindMPE(bnet)
0002 % [mpe_solution,prob_value] =  FindMPE(bnet):
0003 %                           Computes the most probable explanation given the Bayesian network
0004 % INPUTS
0005 % bnet:                Bayesian network
0006 % OUTPUTS
0007 % mpe_solution: Most probable explanation
0008 % prob_value:  Its corresponding probability value
0009 %
0010 % Last version 8/26/2008. Roberto Santana (roberto.santana@ehu.es)
0011 
0012  n = size(bnet.dag,1); % number of nodes
0013  
0014  engine = jtree_inf_engine(bnet, 'maximize', 1);
0015  for i=1:n,
0016   auxvar{i} = [];           %      All variables are hidden
0017  end,
0018  mpe_solution = find_mpe(engine,auxvar);
0019  [eng, logprob] = enter_evidence(engine, mpe_solution);
0020  prob_value = exp(logprob);
0021

Generated on Fri 04-Dec-2009 13:38:29 by m2html © 2003