Mercurial > hg > camir-aes2014
comparison toolboxes/FullBNT-1.0.7/bnt/CPDs/@gaussian_CPD/Old/update_ess.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 CPD = update_ess(CPD, fmarginal, evidence, ns, cnodes, hidden_bitv) | |
2 % UPDATE_ESS Update the Expected Sufficient Statistics of a Gaussian node | |
3 % function CPD = update_ess(CPD, fmarginal, evidence, ns, cnodes, hidden_bitv) | |
4 | |
5 %if nargin < 6 | |
6 % hidden_bitv = zeros(1, max(fmarginal.domain)); | |
7 % hidden_bitv(find(isempty(evidence)))=1; | |
8 %end | |
9 | |
10 dom = fmarginal.domain; | |
11 self = dom(end); | |
12 ps = dom(1:end-1); | |
13 hidden_self = hidden_bitv(self); | |
14 cps = myintersect(ps, cnodes); | |
15 dps = mysetdiff(ps, cps); | |
16 hidden_cps = all(hidden_bitv(cps)); | |
17 hidden_dps = all(hidden_bitv(dps)); | |
18 | |
19 CPD.nsamples = CPD.nsamples + 1; | |
20 [ss cpsz dpsz] = size(CPD.weights); % ss = self size | |
21 | |
22 % Let X be the cts parent (if any), Y be the cts child (self). | |
23 | |
24 if ~hidden_self & (isempty(cps) | ~hidden_cps) & hidden_dps % all cts nodes are observed, all discrete nodes are hidden | |
25 % Since X and Y are observed, SYY = 0, SXX = 0, SXY = 0 | |
26 % Since discrete parents are hidden, we do not need to add evidence to w. | |
27 w = fmarginal.T(:); | |
28 CPD.Wsum = CPD.Wsum + w; | |
29 y = evidence{self}; | |
30 Cyy = y*y'; | |
31 if ~CPD.useC | |
32 W = repmat(w(:)',ss,1); % W(y,i) = w(i) | |
33 W2 = repmat(reshape(W, [ss 1 dpsz]), [1 ss 1]); % W2(x,y,i) = w(i) | |
34 CPD.WYsum = CPD.WYsum + W .* repmat(y(:), 1, dpsz); | |
35 CPD.WYYsum = CPD.WYYsum + W2 .* repmat(reshape(Cyy, [ss ss 1]), [1 1 dpsz]); | |
36 else | |
37 W = w(:)'; | |
38 W2 = reshape(W, [1 1 dpsz]); | |
39 CPD.WYsum = CPD.WYsum + rep_mult(W, y(:), size(CPD.WYsum)); | |
40 CPD.WYYsum = CPD.WYYsum + rep_mult(W2, Cyy, size(CPD.WYYsum)); | |
41 end | |
42 if cpsz > 0 % X exists | |
43 x = cat(1, evidence{cps}); x = x(:); | |
44 Cxx = x*x'; | |
45 Cxy = x*y'; | |
46 if ~CPD.useC | |
47 CPD.WXsum = CPD.WXsum + W .* repmat(x(:), 1, dpsz); | |
48 CPD.WXXsum = CPD.WXXsum + W2 .* repmat(reshape(Cxx, [cpsz cpsz 1]), [1 1 dpsz]); | |
49 CPD.WXYsum = CPD.WXYsum + W2 .* repmat(reshape(Cxy, [cpsz ss 1]), [1 1 dpsz]); | |
50 else | |
51 CPD.WXsum = CPD.WXsum + rep_mult(W, x(:), size(CPD.WXsum)); | |
52 CPD.WXXsum = CPD.WXXsum + rep_mult(W2, Cxx, size(CPD.WXXsum)); | |
53 CPD.WXYsum = CPD.WXYsum + rep_mult(W2, Cxy, size(CPD.WXYsum)); | |
54 end | |
55 end | |
56 return; | |
57 end | |
58 | |
59 % general (non-vectorized) case | |
60 fullm = add_evidence_to_gmarginal(fmarginal, evidence, ns, cnodes); % slow! | |
61 | |
62 if dpsz == 1 % no discrete parents | |
63 w = 1; | |
64 else | |
65 w = fullm.T(:); | |
66 end | |
67 | |
68 CPD.Wsum = CPD.Wsum + w; | |
69 xi = 1:cpsz; | |
70 yi = (cpsz+1):(cpsz+ss); | |
71 for i=1:dpsz | |
72 muY = fullm.mu(yi, i); | |
73 SYY = fullm.Sigma(yi, yi, i); | |
74 CPD.WYsum(:,i) = CPD.WYsum(:,i) + w(i)*muY; | |
75 CPD.WYYsum(:,:,i) = CPD.WYYsum(:,:,i) + w(i)*(SYY + muY*muY'); % E[X Y] = Cov[X,Y] + E[X] E[Y] | |
76 if cpsz > 0 | |
77 muX = fullm.mu(xi, i); | |
78 SXX = fullm.Sigma(xi, xi, i); | |
79 SXY = fullm.Sigma(xi, yi, i); | |
80 CPD.WXsum(:,i) = CPD.WXsum(:,i) + w(i)*muX; | |
81 CPD.WXXsum(:,:,i) = CPD.WXXsum(:,:,i) + w(i)*(SXX + muX*muX'); | |
82 CPD.WXYsum(:,:,i) = CPD.WXYsum(:,:,i) + w(i)*(SXY + muX*muY'); | |
83 end | |
84 end | |
85 |