Daniel@0: function [mpe, prob] = calc_mpe_given_inf_engine(engine, evidence)
Daniel@0: % CALC_MPE_GIVEN_ENGINE Computes the most probable explanation of the evidence
Daniel@0: % [mpe, prob] = calc_mpe_given_inf_engine(engine, evidence)
Daniel@0: %
Daniel@0: % INPUT
Daniel@0: % engine must support max-propagation
Daniel@0: % evidence{i} is the obsevred value of node i, or [] if hidden
Daniel@0: %
Daniel@0: % OUTPUT
Daniel@0: % mpe(i) is the most likely value of node i
Daniel@0: % prob is the likelihood of the globally best assignment
Daniel@0: %
Daniel@0: % This currently only works when all nodes are discrete
Daniel@0: 
Daniel@0: [engine, ll] = enter_evidence(engine, evidence);
Daniel@0: 
Daniel@0: observed = ~isemptycell(evidence);
Daniel@0: N = length(evidence);
Daniel@0: mpe = zeros(1,N);
Daniel@0: for i=1:N
Daniel@0:   m = marginal_nodes(engine, i);
Daniel@0:   % discrete observed nodes are all set to 1 inside the inference engine, so we must undo this
Daniel@0:   if observed(i)
Daniel@0:     mpe(i) = evidence{i};
Daniel@0:   else
Daniel@0:     mpe(i) = argmax(m.T);
Daniel@0:   end
Daniel@0: end
Daniel@0: 
Daniel@0: bnet = bnet_from_engine(engine);
Daniel@0: ll = log_lik_complete(bnet, num2cell(mpe(:)));
Daniel@0: prob = exp(ll);