wolffd@0: function gibbs_test1()
wolffd@0: 
wolffd@0: disp('gibbs test 1')
wolffd@0: 
wolffd@0: rand('state', 0);
wolffd@0: randn('state', 0);
wolffd@0: 
wolffd@0: %[bnet onodes hnodes qnodes] = gibbs_ex_1;
wolffd@0: [bnet onodes hnodes qnodes] = gibbs_ex_2;
wolffd@0: 
wolffd@0: je = jtree_inf_engine(bnet);
wolffd@0: ge = gibbs_sampling_inf_engine (bnet, 'T', 50, 'burnin', 0, ...
wolffd@0: 				'order', [2 2 1 2 1]);
wolffd@0: 
wolffd@0: ev = sample_bnet(bnet);
wolffd@0: 
wolffd@0: evidence = cell(length(bnet.dag), 1);
wolffd@0: evidence(onodes) = ev(onodes);
wolffd@0: [je lj] = enter_evidence(je, evidence);
wolffd@0: [ge lg] = enter_evidence(ge, evidence);
wolffd@0: 
wolffd@0: 
wolffd@0: mj = marginal_nodes(je, qnodes);
wolffd@0: 
wolffd@0: [mg ge] = marginal_nodes (ge, qnodes);
wolffd@0: for t = 1:100
wolffd@0:   [mg ge] = marginal_nodes (ge, qnodes, 'reset_counts', 0);
wolffd@0:   diff = mj.T - mg.T;
wolffd@0:   err(t) = norm (diff(:), 1);
wolffd@0: end
wolffd@0: clf
wolffd@0: plot(err);
wolffd@0: %title('error vs num. Gibbs samples')
wolffd@0: 
wolffd@0: 
wolffd@0: %%%%%%%
wolffd@0: 
wolffd@0: function [bnet, onodes, hnodes, qnodes] = gibbs_ex_1
wolffd@0: % bnet = gibbs_ex_1
wolffd@0: % a simple network to test the gibbs sampling engine
wolffd@0: %    1
wolffd@0: %  / | \
wolffd@0: % 2  3  4
wolffd@0: % |  |  |
wolffd@0: % 5  6  7
wolffd@0: %  \/ \/
wolffd@0: %  8   9
wolffd@0: % where all arcs point downwards
wolffd@0: 
wolffd@0: N = 9;
wolffd@0: dag = zeros(N,N);
wolffd@0: dag(1,2)=1; dag(1,3)=1; dag(1,4)=1;
wolffd@0: dag(2,5)=1; dag(3,6)=1; dag(4,7)=1;
wolffd@0: dag(5,8)=1; dag(6,8)=1; dag(6,9)=1; dag(7,9) = 1;
wolffd@0: 
wolffd@0: onodes = 8:9;
wolffd@0: hnodes = 1:7;
wolffd@0: qnodes = [1 2 6];
wolffd@0: ns = [2 3 4 3 5 2 4 3 2];
wolffd@0: 
wolffd@0: eclass = [1 2 3 2 4 5 6 7 8];
wolffd@0: 
wolffd@0: bnet = mk_bnet (dag, ns, 'equiv_class', eclass);
wolffd@0: 
wolffd@0: for i = 1:3
wolffd@0:   bnet.CPD{i} = tabular_CPD(bnet, i);
wolffd@0: end
wolffd@0: 
wolffd@0: for i = 4:8
wolffd@0:   bnet.CPD{i} = tabular_CPD(bnet, i+1);
wolffd@0: end
wolffd@0: 
wolffd@0: 
wolffd@0: 
wolffd@0: %%%%%%%
wolffd@0: 
wolffd@0: function [bnet, onodes, hnodes, qnodes] = gibbs_ex_2
wolffd@0: % bnet = gibbs_ex_2
wolffd@0: % a very simple network
wolffd@0: %
wolffd@0: % 1   2
wolffd@0: %  \ /
wolffd@0: %   3
wolffd@0: 
wolffd@0: N = 3;
wolffd@0: dag = zeros(N,N);
wolffd@0: dag(1,3)=1; dag(2,3)=1;
wolffd@0: 
wolffd@0: onodes = 3;
wolffd@0: hnodes = 1:2;
wolffd@0: qnodes = 1:2;
wolffd@0: ns = [2 4 3];
wolffd@0: 
wolffd@0: eclass = [1 2 3];
wolffd@0: 
wolffd@0: bnet = mk_bnet (dag, ns, 'equiv_class', eclass);
wolffd@0: 
wolffd@0: for i = 1:3
wolffd@0:   bnet.CPD{i} = tabular_CPD(bnet, i);
wolffd@0: end
wolffd@0: 
wolffd@0: 
wolffd@0: 
wolffd@0: 
wolffd@0: 
wolffd@0: 
wolffd@0: