annotate toolboxes/FullBNT-1.0.7/bnt/general/shrink_obs_dims_in_table.m @ 0:cc4b1211e677
tip
initial commit to HG from
Changeset:
646 (e263d8a21543) added further path and more save "camirversion.m"
| author |
Daniel Wolff |
| date |
Fri, 19 Aug 2016 13:07:06 +0200 |
| parents |
|
| children |
|
| rev |
line source |
|
Daniel@0
|
1 function T2 = shrink_obs_dims_in_table(T1, dom, evidence)
|
|
Daniel@0
|
2 % SHRINK_OBS_DIMS_IN_TABLE Set observed dimensions to size 1
|
|
Daniel@0
|
3 % T2 = shrink_obs_dims_in_table(T1, dom, evidence)
|
|
Daniel@0
|
4 %
|
|
Daniel@0
|
5 % If 'T1' contains observed nodes, it will have 0s in the positions that are
|
|
Daniel@0
|
6 % inconsistent with the evidence. We now remove these 0s and set the corresponding dimensions to
|
|
Daniel@0
|
7 % size 1, to be consistent with the way most inference engines handle evidence, which is to
|
|
Daniel@0
|
8 % shrink observed nodes before doing inference.
|
|
Daniel@0
|
9
|
|
Daniel@0
|
10 % This is used by pearl and enumerative inf. engines.
|
|
Daniel@0
|
11
|
|
Daniel@0
|
12 odom = dom(~isemptycell(evidence(dom)));
|
|
Daniel@0
|
13 vals = cat(1,evidence{odom});
|
|
Daniel@0
|
14 ndx = mk_multi_index(length(dom), find_equiv_posns(odom, dom), vals(:));
|
|
Daniel@0
|
15 T2 = T1(ndx{:});
|
