|
wolffd@0
|
1 function bnet = mk_asia_bnet(CPD_type, p, arity)
|
|
wolffd@0
|
2 % MK_ASIA_BNET Make the 'Asia' bayes net.
|
|
wolffd@0
|
3 %
|
|
wolffd@0
|
4 % BNET = MK_ASIA_BNET uses the parameters specified on p21 of Cowell et al,
|
|
wolffd@0
|
5 % "Probabilistic networks and expert systems", Springer Verlag 1999.
|
|
wolffd@0
|
6 %
|
|
wolffd@0
|
7 % BNET = MK_ASIA_BNET('cpt', p) uses random parameters drawn from a Dirichlet(p,p,...)
|
|
wolffd@0
|
8 % distribution. If p << 1, this is nearly deterministic; if p >> 1, this is nearly uniform.
|
|
wolffd@0
|
9 %
|
|
wolffd@0
|
10 % BNET = MK_ASIA_BNET('bool') makes each CPT a random boolean function.
|
|
wolffd@0
|
11 %
|
|
wolffd@0
|
12 % BNET = MK_ASIA_BNET('gauss') makes each CPT a random linear Gaussian distribution.
|
|
wolffd@0
|
13 %
|
|
wolffd@0
|
14 % BNET = MK_ASIA_BNET('orig') is the same as MK_ASIA_BNET.
|
|
wolffd@0
|
15 %
|
|
wolffd@0
|
16 % BNET = MK_ASIA_BNET('cpt', p, arity) can specify non-binary nodes.
|
|
wolffd@0
|
17
|
|
wolffd@0
|
18
|
|
wolffd@0
|
19 if nargin == 0, CPD_type = 'orig'; end
|
|
wolffd@0
|
20 if nargin < 3, arity = 2; end
|
|
wolffd@0
|
21
|
|
wolffd@0
|
22 Smoking = 1;
|
|
wolffd@0
|
23 Bronchitis = 2;
|
|
wolffd@0
|
24 LungCancer = 3;
|
|
wolffd@0
|
25 VisitToAsia = 4;
|
|
wolffd@0
|
26 TB = 5;
|
|
wolffd@0
|
27 TBorCancer = 6;
|
|
wolffd@0
|
28 Dys = 7;
|
|
wolffd@0
|
29 Xray = 8;
|
|
wolffd@0
|
30
|
|
wolffd@0
|
31 n = 8;
|
|
wolffd@0
|
32 dag = zeros(n);
|
|
wolffd@0
|
33 dag(Smoking, [Bronchitis LungCancer]) = 1;
|
|
wolffd@0
|
34 dag(Bronchitis, Dys) = 1;
|
|
wolffd@0
|
35 dag(LungCancer, TBorCancer) = 1;
|
|
wolffd@0
|
36 dag(VisitToAsia, TB) = 1;
|
|
wolffd@0
|
37 dag(TB, TBorCancer) = 1;
|
|
wolffd@0
|
38 dag(TBorCancer, [Dys Xray]) = 1;
|
|
wolffd@0
|
39
|
|
wolffd@0
|
40 ns = arity*ones(1,n);
|
|
wolffd@0
|
41 if strcmp(CPD_type, 'gauss')
|
|
wolffd@0
|
42 dnodes = [];
|
|
wolffd@0
|
43 else
|
|
wolffd@0
|
44 dnodes = 1:n;
|
|
wolffd@0
|
45 end
|
|
wolffd@0
|
46 bnet = mk_bnet(dag, ns, 'discrete', dnodes);
|
|
wolffd@0
|
47
|
|
wolffd@0
|
48 switch CPD_type
|
|
wolffd@0
|
49 case 'orig',
|
|
wolffd@0
|
50 % true is 2, false is 1
|
|
wolffd@0
|
51 bnet.CPD{VisitToAsia} = tabular_CPD(bnet, VisitToAsia, [0.99 0.01]);
|
|
wolffd@0
|
52 bnet.CPD{Bronchitis} = tabular_CPD(bnet, Bronchitis, [0.7 0.4 0.3 0.6]);
|
|
wolffd@0
|
53 % minka: bug fix
|
|
wolffd@0
|
54 bnet.CPD{Dys} = tabular_CPD(bnet, Dys, [0.9 0.2 0.3 0.1 0.1 0.8 0.7 0.9]);
|
|
wolffd@0
|
55 bnet.CPD{TBorCancer} = tabular_CPD(bnet, TBorCancer, [1 0 0 0 0 1 1 1]);
|
|
wolffd@0
|
56 % minka: bug fix
|
|
wolffd@0
|
57 bnet.CPD{LungCancer} = tabular_CPD(bnet, LungCancer, [0.99 0.9 0.01 0.1]);
|
|
wolffd@0
|
58 bnet.CPD{Smoking} = tabular_CPD(bnet, Smoking, [0.5 0.5]);
|
|
wolffd@0
|
59 bnet.CPD{TB} = tabular_CPD(bnet, TB, [0.99 0.95 0.01 0.05]);
|
|
wolffd@0
|
60 bnet.CPD{Xray} = tabular_CPD(bnet, Xray, [0.95 0.02 0.05 0.98]);
|
|
wolffd@0
|
61 case 'bool',
|
|
wolffd@0
|
62 for i=1:n
|
|
wolffd@0
|
63 bnet.CPD{i} = boolean_CPD(bnet, i, 'rnd');
|
|
wolffd@0
|
64 end
|
|
wolffd@0
|
65 case 'gauss',
|
|
wolffd@0
|
66 for i=1:n
|
|
wolffd@0
|
67 bnet.CPD{i} = gaussian_CPD(bnet, i, 'cov', 1*eye(ns(i)));
|
|
wolffd@0
|
68 end
|
|
wolffd@0
|
69 case 'cpt',
|
|
wolffd@0
|
70 for i=1:n
|
|
wolffd@0
|
71 bnet.CPD{i} = tabular_CPD(bnet, i, p);
|
|
wolffd@0
|
72 end
|
|
wolffd@0
|
73 end
|
|
wolffd@0
|
74
|
|
wolffd@0
|
75
|
|
wolffd@0
|
76
|
