Debugged the thinTree. Now works with tree indexing starting at 1 at the root. TODO : make it work with index 0 on a leaf.

matlab/gtsam_tests/testThinTreeBayesNet.m

@@ -17,7 +17,11 @@
 %  * @author Jean-Guillaume Durand
 %  */
 
+%% Clean the workspace
+clc, clear all, close all;
+
 %% Run the tests
 import gtsam.*
-bayesNet = thinTreeBayesNet(3,2);
+[bayesNet tree] = thinTreeBayesNet(4,2);
 EQUALITY('7 = bayesNet.size', 7, bayesNet.size);
+bayesNet.saveGraph('thinTreeBayesNet.dot');

matlab/gtsam_tests/thinTree.m

@@ -49,12 +49,17 @@ classdef thinTree
         % Function to return the ID's of a node's parents
         function ids = getParents(obj, nodeID)
             % Initialisation
-            ids = zeros(1,obj.w);
             node = nodeID;
-            % Loop on w, the number of parents associated to one node
+            depthOfNode = obj.getNodeDepth(nodeID);
-            for i=1:obj.w
+            if depthOfNode == 1
-                ids(i) = floor(node/2);
+                ids = 1;
-                node = floor(node/2);
+            else
+                ids = zeros(1,min(obj.w, depthOfNode-1));
+                % Loop on w, the number of parents associated to one node
+                for i=1:min(obj.w, depthOfNode-1)
+                    ids(i) = floor(node/2);
+                    node = floor(node/2);
+                end
             end
             % Return
             return
@@ -74,5 +79,10 @@ classdef thinTree
         function output = getNumberOfElements(obj)
             output = 2^obj.depth - 1;
         end
+        
+        % Returns the depth of a node
+        function output = getNodeDepth(obj, nodeID)
+            output = ceil(log(nodeID+1)/log(2));
+        end
     end     % Methods
 end     % Class

matlab/gtsam_tests/thinTreeBayesNet.m

@@ -1,7 +1,7 @@
 function [bayesNet, tree] = thinTreeBayesNet(d,w)
 import gtsam.*
 bayesNet = GaussianBayesNet;
-tree = thinTree(3,2);
+tree = thinTree(d,w);
 
 % Filling the tree
 
@@ -10,24 +10,30 @@ gc = gtsam.GaussianConditional(1, 5*rand(1), 5*rand(1), 3*rand(1));
 % Getting it into the GaussianBayesNet
 bayesNet.push_front(gc);
 
-for i=1:2^tree.getNumberOfElements()
+for i=1:tree.getNumberOfElements()
     % Getting the parents of that node
     parents = tree.getParents(i);
     % Create and link the corresponding GaussianConditionals
-    if tree.getW == 1
+    if tree.getW == 1 || tree.getNodeDepth(i) <= 2
         % Creation of the GaussianConditional
-        gc = gtsam.GaussianConditional(parents(1), 5*rand(1), 5*rand(1));
+        gc = gtsam.GaussianConditional(parents(1), 5*rand(1), 5*rand(1), i, 5*rand(1), 5*rand(1));
         % Getting it into the GaussianBayesNet
         bayesNet.push_front(gc);
         % Getting it in the thinTree
-        t = tree.addContent({gc,parents}, i);
+        tree = tree.addContent({gc,parents}, i);
-    elseif tree.getW == 2
+    elseif tree.getW == 2 && tree.getNodeDepth(i) > 2
-        % Creation of the GaussianConditional
+        % Creation of the GaussianConditional associated with the first
-        gc = gtsam.GaussianConditional(parents(2), 5*rand(1), 5*rand(1), parents(1), 5*rand(1), 5*rand(1));
+        % parent
+        gc = gtsam.GaussianConditional(parents(1), 5*rand(1), 5*rand(1), i, 5*rand(1), 5*rand(1));
+        % Getting it into the GaussianBayesNet
+        bayesNet.push_front(gc);
+        % Creation of the GaussianConditionalj associated with the second
+        % parent
+        gc = gtsam.GaussianConditional(parents(2), 5*rand(1), 5*rand(1), i, 5*rand(1), 5*rand(1));
         % Getting it into the GaussianBayesNet
         bayesNet.push_front(gc);
         % Getting it in the thinTree
-        t = tree.addContent({gc,parents}, i);
+        tree = tree.addContent({gc,parents}, i);
     end
 end
 end