BNW using Octave instead of Matlab.

This version of BNW should perform the same as the original version. The only difference is that it uses Octave instead of Matlab when running BayesNet Toolbox during parameter learning.

I am calling this BNW_1.02. It can be accessed at:
compbio.uthsc.edu/BNW_1.02

1 files changed, 50 insertions, 0 deletions

diff --git a/sourcecodes/bnt-master/KPMtools/marginalize_table.m b/sourcecodes/bnt-master/KPMtools/marginalize_table.m
new file mode 100644
index 00000000..adc5443c
--- /dev/null
+++ b/sourcecodes/bnt-master/KPMtools/marginalize_table.m
@@ -0,0 +1,50 @@
+function smallT = marginalize_table(bigT, bigdom, bigsz, onto, maximize)
+% MARG_TABLE Marginalize a table
+% function smallT = marginalize_table(bigT, bigdom, bigsz, onto, maximize)
+
+% Like marg_table in BNT, except we do not assume the domains are sorted
+
+if nargin < 5, maximize = 0; end
+
+
+smallT = myreshape(bigT, bigsz); % make sure it is a multi-dim array
+sum_over = mysetdiff(bigdom, onto);
+ndx = find_equiv_posns(sum_over, bigdom);
+if maximize
+  for i=1:length(ndx)
+    smallT = max(smallT, [], ndx(i));
+  end
+else
+  for i=1:length(ndx)
+    smallT = sum(smallT, ndx(i));
+  end
+end
+
+
+ns = zeros(1, max(bigdom));
+%ns(bigdom) = mysize(bigT); % ignores trailing dimensions of size 1
+ns(bigdom) = bigsz;
+
+% If onto has a different ordering than bigdom, the following
+% will produce the wrong results
+
+%smallT = squeeze(smallT); % remove all dimensions of size 1
+%smallT = myreshape(smallT, ns(onto)); % put back relevant dims of size 1
+
+% so permute dimensions to match desired ordering (as specified by onto)
+
+
+% like find_equiv_posns, but keeps ordering
+outdom = [onto sum_over];
+for i=1:length(outdom)
+  j = find(bigdom==outdom(i));
+  match(i) = j;
+end
+outdom = [onto sum_over];
+for i=1:length(outdom)
+  j = find(bigdom==outdom(i));
+  match(i) = j;
+end
+if match ~= 1
+  smallT = permute(smallT, match);
+end