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, 43 insertions, 0 deletions

diff --git a/sourcecodes/bnt-master/BNT/inference/static/@pearl_inf_engine/marginal_nodes.m b/sourcecodes/bnt-master/BNT/inference/static/@pearl_inf_engine/marginal_nodes.m
new file mode 100644
index 00000000..bee6ec37
--- /dev/null
+++ b/sourcecodes/bnt-master/BNT/inference/static/@pearl_inf_engine/marginal_nodes.m
@@ -0,0 +1,43 @@
+function marginal = marginal_nodes(engine, query, add_ev)
+% MARGINAL_NODES Compute the marginal on the specified query nodes (loopy)
+% marginal = marginal_nodes(engine, query, add_ev)
+%
+% 'query' must be a single node.
+% add_ev is an optional argument; if 1, observed nodes will be set to their original size,
+% otherwise they will be treated like points.
+   
+if nargin < 3, add_ev = 0; end
+
+if length(query) > 1
+  error('can only compute marginal on single nodes or families')
+end
+bnet = bnet_from_engine(engine);
+ns = bnet.node_sizes(:);
+
+switch engine.msg_type
+ case 'd',
+  T = engine.marginal{query};
+  if ~add_ev
+    marginal.T = shrink_obs_dims_in_table(T, query, engine.evidence);
+  else
+    marginal.T = T;
+  end
+  marginal.domain = query;
+ 
+ case 'g',
+  if engine.disconnected_nodes_bitv(query)
+    marginal.T = 1;
+    marginal.domain = query;
+    if add_ev
+      marginal = add_ev_to_dmarginal(marginal, engine.evidence, ns)
+    end
+    return;
+  end
+
+  marginal = engine.marginal{query};
+  marginal.domain = query;
+  if ~add_ev
+    marginal = shrink_obs_dims_in_gaussian(marginal, query, engine.evidence, ns);
+  end
+end
+