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, 24 insertions, 0 deletions

diff --git a/sourcecodes/bnt-master/KPMtools/logsumexp.m b/sourcecodes/bnt-master/KPMtools/logsumexp.m
new file mode 100644
index 00000000..b639cbb2
--- /dev/null
+++ b/sourcecodes/bnt-master/KPMtools/logsumexp.m
@@ -0,0 +1,24 @@
+function s = logsumexp(a, dim)
+% Returns log(sum(exp(a),dim)) while avoiding numerical underflow.
+% Default is dim = 1 (rows) or dim=2 for a row vector
+% logsumexp(a, 2) will sum across columns instead of rows
+
+% Written by Tom Minka, modified by Kevin Murphy
+
+if nargin < 2
+  dim = 1;
+  if ndims(a) <= 2 & size(a,1)==1
+    dim = 2;
+  end
+end
+
+% subtract the largest in each column
+[y, i] = max(a,[],dim);
+dims = ones(1,ndims(a));
+dims(dim) = size(a,dim);
+a = a - repmat(y, dims);
+s = y + log(sum(exp(a),dim));
+%i = find(~finite(y));
+%if ~isempty(i)
+%  s(i) = y(i);
+%end