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, 30 insertions, 0 deletions

diff --git a/sourcecodes/bnt-master/BNT/general/fgraph_to_bnet.m b/sourcecodes/bnt-master/BNT/general/fgraph_to_bnet.m
new file mode 100644
index 00000000..9aaf40ef
--- /dev/null
+++ b/sourcecodes/bnt-master/BNT/general/fgraph_to_bnet.m
@@ -0,0 +1,30 @@
+function bnet = fgraph_to_bnet(fg)
+% FGRAPH_TO_BNET Convert a factor graph to a Bayes net
+% bnet = fgraph_to_bnet(fg)
+%
+% We assume all factors are tabular_CPD.
+% We create 1 dummy observed node for every factor.
+
+N = fg.nvars + fg.nfactors;
+vnodes = 1:fg.nvars;
+fnodes = fg.nvars+1:N;
+dag = zeros(N);
+for x=1:fg.nvars
+  dag(x, fnodes(fg.dep{x})) = 1;
+end
+ns = [fg.node_sizes ones(1, fg.nfactors)];
+discrete = [fg.dnodes fnodes];
+bnet = mk_bnet(dag, ns, 'discrete', discrete);
+for x=1:fg.nvars
+  bnet.CPD{x} = tabular_CPD(bnet, x, 'CPT', 'unif');
+end
+ev = cell(1, fg.nvars); % no evidence
+for i=1:fg.nfactors
+  f = fnodes(i);
+  e = fg.equiv_class(i);
+  pot = convert_to_pot(fg.factors{e}, 'd', fg.dom{i}, ev);
+  m = pot_to_marginal(pot);
+  bnet.CPD{f} = tabular_CPD(bnet, f, 'CPT', m.T);
+end
+  
+