|
wolffd@0
|
1 function bnet = mk_bnet(dag, node_sizes, varargin)
|
|
wolffd@0
|
2 % MK_BNET Make a Bayesian network.
|
|
wolffd@0
|
3 %
|
|
wolffd@0
|
4 % BNET = MK_BNET(DAG, NODE_SIZES, ...) makes a graphical model with an arc from i to j iff DAG(i,j) = 1.
|
|
wolffd@0
|
5 % Thus DAG is the adjacency matrix for a directed acyclic graph.
|
|
wolffd@0
|
6 % The nodes are assumed to be in topological order. Use TOPOLOGICAL_SORT if necessary.
|
|
wolffd@0
|
7 %
|
|
wolffd@0
|
8 % node_sizes(i) is the number of values node i can take on,
|
|
wolffd@0
|
9 % or the length of node i if i is a continuous-valued vector.
|
|
wolffd@0
|
10 % node_sizes(i) = 1 if i is a utility node.
|
|
wolffd@0
|
11 %
|
|
wolffd@0
|
12 % Below are the names of optional arguments [and their default value in brackets].
|
|
wolffd@0
|
13 % Pass as 'PropertyName1', PropertyValue1, 'PropertyName2', PropertyValue2, ...
|
|
wolffd@0
|
14 %
|
|
wolffd@0
|
15 % discrete - the list of nodes which are discrete random variables [1:N]
|
|
wolffd@0
|
16 % equiv_class - equiv_class(i)=j means node i gets its params from CPD{j} [1:N]
|
|
wolffd@0
|
17 % observed - the list of nodes which will definitely be observed in every case [ [] ]
|
|
wolffd@0
|
18 % 'names' - a cell array of strings to be associated with nodes 1:n [{}]
|
|
wolffd@0
|
19 % This creates an associative array, so you write e.g.
|
|
wolffd@0
|
20 % 'evidence(bnet.names{'bar'}) = 42' instead of 'evidence(2} = 42'
|
|
wolffd@0
|
21 % assuming names = { 'foo', 'bar', ...}.
|
|
wolffd@0
|
22 %
|
|
wolffd@0
|
23 % e.g., bnet = mk_bnet(dag, ns, 'discrete', [1 3])
|
|
wolffd@0
|
24 %
|
|
wolffd@0
|
25 % For backwards compatibility with BNT2, you can also specify the parameters in the following order
|
|
wolffd@0
|
26 % bnet = mk_bnet(dag, node_sizes, discrete_nodes, equiv_class)
|
|
wolffd@0
|
27
|
|
wolffd@0
|
28 n = length(dag);
|
|
wolffd@0
|
29
|
|
wolffd@0
|
30 % default values for parameters
|
|
wolffd@0
|
31 bnet.equiv_class = 1:n;
|
|
wolffd@0
|
32 bnet.dnodes = 1:n; % discrete
|
|
wolffd@0
|
33 bnet.observed = [];
|
|
wolffd@0
|
34 bnet.names = {};
|
|
wolffd@0
|
35
|
|
wolffd@0
|
36 if nargin >= 3
|
|
wolffd@0
|
37 args = varargin;
|
|
wolffd@0
|
38 nargs = length(args);
|
|
wolffd@0
|
39 if ~isstr(args{1})
|
|
wolffd@0
|
40 if nargs >= 1, bnet.dnodes = args{1}; end
|
|
wolffd@0
|
41 if nargs >= 2, bnet.equiv_class = args{2}; end
|
|
wolffd@0
|
42 else
|
|
wolffd@0
|
43 for i=1:2:nargs
|
|
wolffd@0
|
44 switch args{i},
|
|
wolffd@0
|
45 case 'equiv_class', bnet.equiv_class = args{i+1};
|
|
wolffd@0
|
46 case 'discrete', bnet.dnodes = args{i+1};
|
|
wolffd@0
|
47 case 'observed', bnet.observed = args{i+1};
|
|
wolffd@0
|
48 case 'names', bnet.names = assocarray(args{i+1}, num2cell(1:n));
|
|
wolffd@0
|
49 otherwise,
|
|
wolffd@0
|
50 error(['invalid argument name ' args{i}]);
|
|
wolffd@0
|
51 end
|
|
wolffd@0
|
52 end
|
|
wolffd@0
|
53 end
|
|
wolffd@0
|
54 end
|
|
wolffd@0
|
55
|
|
wolffd@0
|
56 bnet.observed = sort(bnet.observed); % for comparing sets
|
|
wolffd@0
|
57 bnet.hidden = mysetdiff(1:n, bnet.observed(:)');
|
|
wolffd@0
|
58 bnet.hidden_bitv = zeros(1,n);
|
|
wolffd@0
|
59 bnet.hidden_bitv(bnet.hidden) = 1;
|
|
wolffd@0
|
60 bnet.dag = dag;
|
|
wolffd@0
|
61 bnet.node_sizes = node_sizes(:)';
|
|
wolffd@0
|
62
|
|
wolffd@0
|
63 bnet.cnodes = mysetdiff(1:n, bnet.dnodes);
|
|
wolffd@0
|
64 % too many functions refer to cnodes to rename it to cts_nodes -
|
|
wolffd@0
|
65 % We hope it won't be confused with chance nodes!
|
|
wolffd@0
|
66
|
|
wolffd@0
|
67 bnet.parents = cell(1,n);
|
|
wolffd@0
|
68 for i=1:n
|
|
wolffd@0
|
69 bnet.parents{i} = parents(dag, i);
|
|
wolffd@0
|
70 end
|
|
wolffd@0
|
71
|
|
wolffd@0
|
72 E = max(bnet.equiv_class);
|
|
wolffd@0
|
73 mem = cell(1,E);
|
|
wolffd@0
|
74 for i=1:n
|
|
wolffd@0
|
75 e = bnet.equiv_class(i);
|
|
wolffd@0
|
76 mem{e} = [mem{e} i];
|
|
wolffd@0
|
77 end
|
|
wolffd@0
|
78 bnet.members_of_equiv_class = mem;
|
|
wolffd@0
|
79
|
|
wolffd@0
|
80 bnet.CPD = cell(1, E);
|
|
wolffd@0
|
81
|
|
wolffd@0
|
82 bnet.rep_of_eclass = zeros(1,E);
|
|
wolffd@0
|
83 for e=1:E
|
|
wolffd@0
|
84 mems = bnet.members_of_equiv_class{e};
|
|
wolffd@0
|
85 bnet.rep_of_eclass(e) = mems(1);
|
|
wolffd@0
|
86 end
|
|
wolffd@0
|
87
|
|
wolffd@0
|
88 directed = 1;
|
|
wolffd@0
|
89 if ~acyclic(dag,directed)
|
|
wolffd@0
|
90 error('graph must be acyclic')
|
|
wolffd@0
|
91 end
|
|
wolffd@0
|
92
|
|
wolffd@0
|
93 bnet.order = topological_sort(bnet.dag);
|
