wolffd@0
|
1 function mpe = find_mpe(engine, evidence, varargin)
|
wolffd@0
|
2 % FIND_MPE Find the most probable explanation of the data (assignment to the hidden nodes)
|
wolffd@0
|
3 % function mpe = find_mpe(engine, evidence,...)
|
wolffd@0
|
4 %
|
wolffd@0
|
5 % evidence{i} = [] if X(i) is hidden, and otherwise contains its observed value (scalar or column vector).
|
wolffd@0
|
6 %
|
wolffd@0
|
7 % The following optional arguments can be specified in the form of name/value pairs:
|
wolffd@0
|
8 % [default value in brackets]
|
wolffd@0
|
9 %
|
wolffd@0
|
10 % soft - a cell array of soft/virtual evidence;
|
wolffd@0
|
11 % soft{i} is a prob. distrib. over i's values, or [] [ cell(1,N) ]
|
wolffd@0
|
12 %
|
wolffd@0
|
13
|
wolffd@0
|
14 bnet = bnet_from_engine(engine);
|
wolffd@0
|
15 ns = bnet.node_sizes(:);
|
wolffd@0
|
16 N = length(bnet.dag);
|
wolffd@0
|
17
|
wolffd@0
|
18 engine.evidence = evidence;
|
wolffd@0
|
19
|
wolffd@0
|
20 % set default params
|
wolffd@0
|
21 exclude = [];
|
wolffd@0
|
22 soft_evidence = cell(1,N);
|
wolffd@0
|
23
|
wolffd@0
|
24 % parse optional params
|
wolffd@0
|
25 args = varargin;
|
wolffd@0
|
26 nargs = length(args);
|
wolffd@0
|
27 for i=1:2:nargs
|
wolffd@0
|
28 switch args{i},
|
wolffd@0
|
29 case 'soft', soft_evidence = args{i+1};
|
wolffd@0
|
30 otherwise,
|
wolffd@0
|
31 error(['invalid argument name ' args{i}]);
|
wolffd@0
|
32 end
|
wolffd@0
|
33 end
|
wolffd@0
|
34 engine.maximize = 1;
|
wolffd@0
|
35
|
wolffd@0
|
36 onodes = find(~isemptycell(evidence));
|
wolffd@0
|
37 hnodes = find(isemptycell(evidence));
|
wolffd@0
|
38 pot_type = determine_pot_type(bnet, onodes);
|
wolffd@0
|
39 if strcmp(pot_type, 'cg')
|
wolffd@0
|
40 check_for_cd_arcs(onodes, bnet.cnodes, bnet.dag);
|
wolffd@0
|
41 end
|
wolffd@0
|
42
|
wolffd@0
|
43 hard_nodes = 1:N;
|
wolffd@0
|
44 soft_nodes = find(~isemptycell(soft_evidence));
|
wolffd@0
|
45 S = length(soft_nodes);
|
wolffd@0
|
46 if S > 0
|
wolffd@0
|
47 assert(pot_type == 'd');
|
wolffd@0
|
48 assert(mysubset(soft_nodes, bnet.dnodes));
|
wolffd@0
|
49 end
|
wolffd@0
|
50
|
wolffd@0
|
51 % Evaluate CPDs with evidence, and convert to potentials
|
wolffd@0
|
52 pot = cell(1, N+S);
|
wolffd@0
|
53 for n=1:N
|
wolffd@0
|
54 fam = family(bnet.dag, n);
|
wolffd@0
|
55 e = bnet.equiv_class(n);
|
wolffd@0
|
56 if isempty(bnet.CPD{e})
|
wolffd@0
|
57 error(['must define CPD ' num2str(e)])
|
wolffd@0
|
58 else
|
wolffd@0
|
59 pot{n} = convert_to_pot(bnet.CPD{e}, pot_type, fam(:), evidence);
|
wolffd@0
|
60 end
|
wolffd@0
|
61 end
|
wolffd@0
|
62
|
wolffd@0
|
63 for i=1:S
|
wolffd@0
|
64 n = soft_nodes(i);
|
wolffd@0
|
65 pot{N+i} = dpot(n, ns(n), soft_evidence{n});
|
wolffd@0
|
66 end
|
wolffd@0
|
67 clqs = engine.clq_ass_to_node([hard_nodes soft_nodes]);
|
wolffd@0
|
68
|
wolffd@0
|
69 [clpot, seppot] = init_pot(engine, clqs, pot, pot_type, onodes);
|
wolffd@0
|
70 [clpot, seppot] = collect_evidence(engine, clpot, seppot);
|
wolffd@0
|
71 mpe = find_max_config(engine, clpot, seppot);
|