[oMetah-devel] ometah/metaheuristic itsEstimationOfDistributio...

[NoJhan]

CVSROOT:        /cvsroot/ometah
Module name:    ometah
Branch:         
Changes by:     NoJhan <address@hidden> 05/06/12 16:11:13

Modified files:
        metaheuristic  : itsEstimationOfDistribution.cpp 
                         itsEstimationOfDistribution.hpp 

Log message:
        * adding a normal probability density function

CVSWeb URLs:
http://savannah.gnu.org/cgi-bin/viewcvs/ometah/ometah/metaheuristic/itsEstimationOfDistribution.cpp.diff?tr1=1.17&tr2=1.18&r1=text&r2=text
http://savannah.gnu.org/cgi-bin/viewcvs/ometah/ometah/metaheuristic/itsEstimationOfDistribution.hpp.diff?tr1=1.9&tr2=1.10&r1=text&r2=text

Patches:
Index: ometah/metaheuristic/itsEstimationOfDistribution.cpp
diff -u ometah/metaheuristic/itsEstimationOfDistribution.cpp:1.17 
ometah/metaheuristic/itsEstimationOfDistribution.cpp:1.18
--- ometah/metaheuristic/itsEstimationOfDistribution.cpp:1.17   Tue Jun  7 
12:11:57 2005
+++ ometah/metaheuristic/itsEstimationOfDistribution.cpp        Sun Jun 12 
16:11:13 2005
@@ -1,5 +1,5 @@
 /***************************************************************************
- *  $Id: itsEstimationOfDistribution.cpp,v 1.17 2005/06/07 12:11:57 nojhan Exp 
$
+ *  $Id: itsEstimationOfDistribution.cpp,v 1.18 2005/06/12 16:11:13 nojhan Exp 
$
  *  Copyright : Université Paris 12 Val-de-Marne
  *              (61 avenue du Général de Gaulle, 94010, Créteil, France)
  *  Author : Johann Dréo <address@hidden>
@@ -65,25 +65,19 @@
     setFamily("Estimation of Distribution Algorithm");
     
 
-    setDistribution("Uniform");
+    //setDistribution("Uniform");
+    setDistribution("Normal");
+    
     setSelectRatio(0.5);
 }
 
 
-void itsEstimationOfDistribution::learning()
+void itsEstimationOfDistribution::learningUniform(vector< vector<double> > 
aSample)
 {
-    vector< vector<double> > transposedSimplifiedSample;
-    
-    // for each point, extract the solution vector
-    for(unsigned int i=0; i < getSampleSizeCurrent(); i++) {
-        // put this vector in a matrix
-        transposedSimplifiedSample.push_back( sample[i].getSolution() );
-    }
-
     // transpose it so that the row correspond to a dimension, and not a 
solution
-    transposedSimplifiedSample = transpose(transposedSimplifiedSample);
+    vector< vector<double> > transposedSimplifiedSample = transpose(aSample);
 
-// verify that we have the correct dimension
+    // verify that we have the correct dimension
     if( transposedSimplifiedSample.size() != (unsigned 
int)this->problem->getDimension() ) {
         ostringstream err;
         err << "ErrorSize: transposed matrix size (" 
@@ -99,8 +93,40 @@
     this->parameterUniformMaxima.clear();
     for(int dim=0; dim < this->problem->getDimension(); dim++) {
         // store them
-        this->parameterUniformMinima.push_back( min( 
transposedSimplifiedSample[dim] ) );
-        this->parameterUniformMaxima.push_back( max( 
transposedSimplifiedSample[dim] ) );
+        this->parameterUniformMinima.push_back( min( aSample[dim] ) );
+        this->parameterUniformMaxima.push_back( max( aSample[dim] ) );
+    }
+}
+
+void itsEstimationOfDistribution::learningNormal(vector< vector<double> > 
aSample)
+{
+    //! mean on each dimension
+    this->parameterNormalMean = mean( transpose(aSample) );
+
+    //! variance covariance matrix
+    this->parameterNormalVarCovar = varianceCovariance(aSample);
+}
+
+
+void itsEstimationOfDistribution::learning()
+{
+    vector< vector<double> > simplifiedSample;
+    
+    // for each point, extract the solution vector
+    for(unsigned int i=0; i < getSampleSizeCurrent(); i++) {
+        // put this vector in a matrix
+        simplifiedSample.push_back( sample[i].getSolution() );
+    }
+
+
+    if ( distribution == "Uniform" ) {
+        learningUniform( simplifiedSample );
+    
+    } else if (distribution == "Normal") {  
+        learningNormal( simplifiedSample );
+    
+    } else {
+        learningUniform( simplifiedSample );
     }
 }
  
@@ -113,7 +139,17 @@
     for( unsigned int i=0; i < getSampleSize(); i++) {
         // draw solution
         itsPoint p;
-        p.setSolution( randomUniform(this->parameterUniformMinima, 
this->parameterUniformMaxima) );
+    
+        if ( distribution == "Uniform" ) {
+            p.setSolution( randomUniform(this->parameterUniformMinima, 
this->parameterUniformMaxima) );
+        
+        } else if (distribution == "Normal") {  
+            p.setSolution( randomNormalMulti(this->parameterNormalMean, 
this->parameterNormalVarCovar) );
+        
+        } else {
+            p.setSolution( randomUniform(this->parameterUniformMinima, 
this->parameterUniformMaxima) );
+        }
+    
         // get values
         sample[i] = evaluate(p);
     }
Index: ometah/metaheuristic/itsEstimationOfDistribution.hpp
diff -u ometah/metaheuristic/itsEstimationOfDistribution.hpp:1.9 
ometah/metaheuristic/itsEstimationOfDistribution.hpp:1.10
--- ometah/metaheuristic/itsEstimationOfDistribution.hpp:1.9    Tue Jun  7 
12:11:57 2005
+++ ometah/metaheuristic/itsEstimationOfDistribution.hpp        Sun Jun 12 
16:11:13 2005
@@ -1,5 +1,5 @@
 /***************************************************************************
- *  $Id: itsEstimationOfDistribution.hpp,v 1.9 2005/06/07 12:11:57 nojhan Exp $
+ *  $Id: itsEstimationOfDistribution.hpp,v 1.10 2005/06/12 16:11:13 nojhan Exp 
$
  *  Copyright : Université Paris 12 Val-de-Marne
  *              (61 avenue du Général de Gaulle, 94010, Créteil, France)
  *  Author : Johann Dréo <address@hidden>
@@ -37,6 +37,7 @@
   /*!
     Either :
       - Uniform
+      - Normal
   */
   string distribution;
 
@@ -46,6 +47,12 @@
   //! The maxima parameter for the uniform distribution
   vector<double> parameterUniformMaxima;
 
+  //! The minima parameter for the uniform distribution
+  vector<double> parameterNormalMean;
+
+  //! The maxima parameter for the uniform distribution
+  vector<vector<double> > parameterNormalVarCovar;
+
   //! The select ratio
   /*!
     A number in [0,1] telling what proportion of points
@@ -63,6 +70,11 @@
   //! the learning
   void learning();
       
+  //! Learn the parameters of an uniform distribution
+  void learningUniform(vector< vector<double> > aSample);
+
+  //! Learn the parameters of a normal distribution
+  void learningNormal(vector< vector<double> > aSample);
 
 public: