wolffd@0
|
1 % Compare various inference engines on the following network (from Jensen (1996) p84 fig 4.17)
|
wolffd@0
|
2 % 1
|
wolffd@0
|
3 % / | \
|
wolffd@0
|
4 % 2 3 4
|
wolffd@0
|
5 % | | |
|
wolffd@0
|
6 % 5 6 7
|
wolffd@0
|
7 % \/ \/
|
wolffd@0
|
8 % 8 9
|
wolffd@0
|
9 % where all arcs point downwards
|
wolffd@0
|
10
|
wolffd@0
|
11 N = 9;
|
wolffd@0
|
12 dag = zeros(N,N);
|
wolffd@0
|
13 dag(1,2)=1; dag(1,3)=1; dag(1,4)=1;
|
wolffd@0
|
14 dag(2,5)=1; dag(3,6)=1; dag(4,7)=1;
|
wolffd@0
|
15 dag(5,8)=1; dag(6,8)=1; dag(6,9)=1; dag(7,9) = 1;
|
wolffd@0
|
16
|
wolffd@0
|
17 dnodes = 1:N;
|
wolffd@0
|
18 false = 1; true = 2;
|
wolffd@0
|
19 ns = 2*ones(1,N); % binary nodes
|
wolffd@0
|
20
|
wolffd@0
|
21 onodes = [1];
|
wolffd@0
|
22 evidence = cell(1,N);
|
wolffd@0
|
23 evidence(onodes) = num2cell(1);
|
wolffd@0
|
24 bnet = mk_bnet(dag, ns, 'observed', onodes);
|
wolffd@0
|
25 % use random params
|
wolffd@0
|
26 %for i=1:N
|
wolffd@0
|
27 % bnet.CPD{i} = tabular_CPD(bnet, i);
|
wolffd@0
|
28 %end
|
wolffd@0
|
29 bnet.CPD{1} = tabular_CPD(bnet, 1, 'sparse', 1, 'CPT', [0.8, 0.2]);
|
wolffd@0
|
30 bnet.CPD{2} = tabular_CPD(bnet, 2, 'sparse', 1, 'CPT', [1 0 0 1]);
|
wolffd@0
|
31 bnet.CPD{3} = tabular_CPD(bnet, 3, 'sparse', 1, 'CPT', [0 1 1 0]);
|
wolffd@0
|
32 bnet.CPD{4} = tabular_CPD(bnet, 4, 'sparse', 1, 'CPT', [1 1 0 0]);
|
wolffd@0
|
33 bnet.CPD{5} = tabular_CPD(bnet, 5, 'sparse', 1, 'CPT', [0 0 1 1]);
|
wolffd@0
|
34 bnet.CPD{6} = tabular_CPD(bnet, 6, 'sparse', 1, 'CPT', [1 0 0 1]);
|
wolffd@0
|
35 bnet.CPD{7} = tabular_CPD(bnet, 7, 'sparse', 1, 'CPT', [0 1 1 0]);
|
wolffd@0
|
36 bnet.CPD{8} = tabular_CPD(bnet, 8, 'sparse', 1, 'CPT', [1 1 0 0 0 0 1 1]);
|
wolffd@0
|
37 bnet.CPD{9} = tabular_CPD(bnet, 9, 'sparse', 1, 'CPT', [0 1 0 1 1 0 1 0]);
|
wolffd@0
|
38
|
wolffd@0
|
39 engine = jtree_sparse_inf_engine(bnet);
|
wolffd@0
|
40 tic
|
wolffd@0
|
41 [engine, ll] = enter_evidence(engine, evidence);
|
wolffd@0
|
42 toc
|
wolffd@0
|
43
|