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, 117 insertions, 0 deletions

diff --git a/sourcecodes/bnt-master/GraphViz/draw_hmm.m b/sourcecodes/bnt-master/GraphViz/draw_hmm.m
new file mode 100644
index 00000000..e517789c
--- /dev/null
+++ b/sourcecodes/bnt-master/GraphViz/draw_hmm.m
@@ -0,0 +1,117 @@
+function draw_hmm(A, varargin)
+% DRAW_HMM Make a picture of the HMM using dotty
+% function draw_hmm(A, ...)
+%
+% For details on dotty, see http://www.research.att.com/sw/tools/graphviz
+%
+% If A(i,j) > thresh, we draw and arc from state i to state j.
+%
+% Optional arguments (name/value pairs) [default]
+%
+% thresh - [1e-1]
+% obsprob - If B(i,o) > 0, we include "o" in the name of state i.
+%     e.g., if state 5 emits 1,3,7, its label becomes "5: 1 3 7".
+% startprob - ifstartprob(i) > 0, the state name will be prefixed with "+".
+% endprob - if endprob(i) > 0, the state name will be appended with "-".
+% filename - if [], we write to 'tmp.dot', convert this to 'tmp.ps'
+%  using 'dot -Tps tmp.dot -o tmp.ps', and then call ghostview  to display the result.
+%  dot and gv must be on your system path.
+%  If filename ~= [], we just generate the dot file, and do not
+%  convert it to postscript or call ghostview.
+
+[thresh, B, startprob, endprob, filename] = ...
+    process_options(varargin, 'thresh', 1e-1, 'obsprob', [], 'startprob', [], 'endprob', [], ...
+		    'filename', []);
+
+Q = length(A);
+
+arclabel = cell(Q,Q);
+G = zeros(Q,Q);
+for i=1:Q
+  for j=1:Q
+    if A(i,j) < thresh
+      arclabel{i,j} = '';
+    else
+      G(i,j) = 1;
+      arclabel{i,j} = sprintf('%5.3f', A(i,j));
+    end
+  end
+end
+
+
+nodelabel = cell(1,Q);
+for i=1:Q
+  % annotate start/stop states
+  if ~isempty(startprob) & ~approxeq(startprob(i), 0)
+    start = '+';
+  else
+    start = '';
+  end
+  if ~isempty(endprob) & ~approxeq(hmm.endprob(i), 0)
+    stop = '-';
+  else
+    stop = '';
+  end
+  label = sprintf('%s%d%s :', start, i, stop);
+
+  if ~isempty(B)
+    output_label = mk_output_label(B);
+    label = strcat(label, output_label);
+  end
+  
+  nodelabel{i} = label;
+end
+
+
+if isempty(filename)
+  filename = 'tmp.dot';
+  %mkdot(G, filename, arclabel, nodelabel)  
+  graph_to_dot(G, 'filename', filename, 'arc_label', arclabel, 'node_label', nodelabel);
+  fprintf('converting from .ps to .dot\n')
+  !dot -Tps tmp.dot -o tmp.ps
+  !gv tmp.ps &
+else
+  graph_to_dot(G, 'filename', filename, 'arc_label', arclabel, 'node_label', nodelabel);
+  %mkdot(G, filename, arclabel, nodelabel)  
+end
+
+
+%%%%%%%%%
+
+function label = mk_output_label(B)
+
+[Q O] = size(B);
+label = '';
+
+if 0
+  % print most probable symbols
+  for i=1:Q
+    m = max(B(i,:));
+    ndx = find(abs(B(i,:) - repmat(m,1,O)) < 1e-2);
+    %ndx = find(B(i,:)==m);
+    %label = sprintf('%d,', ndx);
+  end
+end
+
+if 0
+  % print prob distrib over all symbols 
+  for o=1:O
+    if approxeq(B(i,o), 0)
+      %
+    else
+      label = strcat(label, sprintf('%d(%3.2f),', o, B(i,o)));
+    end
+  end
+end
+
+if 1
+  % print all non-zero symbols
+  chars = ['a' 'b' 'c'];
+  for o=1:O
+    if approxeq(B(i,o), 0)
+      %
+    else
+      label = strcat(label, sprintf('%s', chars(o)));
+    end
+  end
+end