Mercurial > hg > camir-aes2014
comparison toolboxes/FullBNT-1.0.7/bnt/CPDs/@tabular_CPD/log_marg_prob_node.m @ 0:e9a9cd732c1e tip
first hg version after svn
author | wolffd |
---|---|
date | Tue, 10 Feb 2015 15:05:51 +0000 |
parents | |
children |
comparison
equal
deleted
inserted
replaced
-1:000000000000 | 0:e9a9cd732c1e |
---|---|
1 function L = log_marg_prob_node(CPD, self_ev, pev, usecell) | |
2 % LOG_MARG_PROB_NODE Compute sum_m log P(x(i,m)| x(pi_i,m)) for node i (tabular) | |
3 % L = log_marg_prob_node(CPD, self_ev, pev) | |
4 % | |
5 % This differs from log_prob_node because we integrate out the parameters. | |
6 % self_ev(m) is the evidence on this node in case m. | |
7 % pev(i,m) is the evidence on the i'th parent in case m (if there are any parents). | |
8 % (These may also be cell arrays.) | |
9 | |
10 ncases = length(self_ev); | |
11 sz = CPD.sizes; | |
12 nparents = length(sz)-1; | |
13 assert(ncases == size(pev, 2)); | |
14 | |
15 if nargin < 4 | |
16 %usecell = 0; | |
17 if iscell(self_ev) | |
18 usecell = 1; | |
19 else | |
20 usecell = 0; | |
21 end | |
22 end | |
23 | |
24 | |
25 if ncases==0 | |
26 L = 0; | |
27 return; | |
28 elseif ncases==1 % speedup the sequential learning case | |
29 CPT = CPD.CPT; | |
30 % We assume the CPTs are already set to the mean of the posterior (due to bayes_update_params) | |
31 if usecell | |
32 x = cat(1, pev{:})'; | |
33 y = self_ev{1}; | |
34 else | |
35 %x = pev(:)'; | |
36 x = pev; | |
37 y = self_ev; | |
38 end | |
39 switch nparents | |
40 case 0, p = CPT(y); | |
41 case 1, p = CPT(x(1), y); | |
42 case 2, p = CPT(x(1), x(2), y); | |
43 case 3, p = CPT(x(1), x(2), x(3), y); | |
44 otherwise, | |
45 ind = subv2ind(sz, [x y]); | |
46 p = CPT(ind); | |
47 end | |
48 L = log(p); | |
49 else | |
50 % We ignore the CPTs here and assume the prior has not been changed | |
51 | |
52 % We arrange the data as in the following example. | |
53 % Let there be 2 parents and 3 cases. Let p(i,m) be parent i in case m, | |
54 % and y(m) be the child in case m. Then we create the data matrix | |
55 % | |
56 % p(1,1) p(1,2) p(1,3) | |
57 % p(2,1) p(2,2) p(2,3) | |
58 % y(1) y(2) y(3) | |
59 if usecell | |
60 data = [cell2num(pev); cell2num(self_ev)]; | |
61 else | |
62 data = [pev; self_ev]; | |
63 end | |
64 %S = struct(CPD); fprintf('log marg prob node %d, ps\n', S.self); disp(S.parents) | |
65 counts = compute_counts(data, sz); | |
66 L = dirichlet_score_family(counts, CPD.dirichlet); | |
67 end | |
68 | |
69 |