Daniel@0: % This example is from Page.143 of "Probabilistic Networks and Expert Systems",
Daniel@0: % Cowell, Dawid, Lauritzen and Spiegelhalter, 1999, Springer.
Daniel@0: 
Daniel@0: X = 1; Y = 2; Z = 3;
Daniel@0: n = 3;
Daniel@0: 
Daniel@0: dag = zeros(n);
Daniel@0: dag(X, Y)=1;
Daniel@0: dag(Y, Z)=1;
Daniel@0: 
Daniel@0: ns = ones(1, n);
Daniel@0: dnodes = [];
Daniel@0: 
Daniel@0: bnet = mk_bnet(dag, ns, dnodes);
Daniel@0: bnet.CPD{X} = gaussian_CPD(bnet, X, 'mean', 0, 'cov', 1);
Daniel@0: bnet.CPD{Y} = gaussian_CPD(bnet, Y, 'mean', 0, 'cov', 1, 'weights', 1);
Daniel@0: bnet.CPD{Z} = gaussian_CPD(bnet, Z, 'mean', 0, 'cov', 1, 'weights', 1);
Daniel@0: 
Daniel@0: engines = {};
Daniel@0: engines{end+1} = jtree_inf_engine(bnet);
Daniel@0: engines{end+1} = stab_cond_gauss_inf_engine(bnet);
Daniel@0: nengines = length(engines);
Daniel@0: 
Daniel@0: evidence = cell(1,n);
Daniel@0: evidence{Y} = 1.5; 
Daniel@0: 
Daniel@0: for e=1:nengines
Daniel@0:   engines{e} = enter_evidence(engines{e}, evidence);
Daniel@0:   margX = marginal_nodes(engines{e}, X);
Daniel@0:   assert(approxeq(margX.mu, 0.75))
Daniel@0:   assert(approxeq(margX.Sigma, 0.5))
Daniel@0:   
Daniel@0:   margZ = marginal_nodes(engines{e}, Z);
Daniel@0:   assert(approxeq(margZ.mu, 1.5))
Daniel@0:   assert(approxeq(margZ.Sigma, 1))
Daniel@0: end
Daniel@0: 
Daniel@0: 
Daniel@0: evidence = cell(1,n);
Daniel@0: evidence{Z} = 1.5; 
Daniel@0: 
Daniel@0: for e=1:nengines
Daniel@0:   engines{e} = enter_evidence(engines{e}, evidence);
Daniel@0:   margX = marginal_nodes(engines{e}, X);
Daniel@0:   assert(approxeq(margX.mu, 1/2))
Daniel@0:   assert(approxeq(margX.Sigma, 2/3))
Daniel@0:   
Daniel@0:   margY = marginal_nodes(engines{e}, Y);
Daniel@0:   assert(approxeq(margY.mu, 1))
Daniel@0:   assert(approxeq(margY.Sigma, 2/3))
Daniel@0: end
Daniel@0: