Ajout des statistics, j'espère avoir fini ;)

git-svn-id: svn://scm.gforge.inria.fr/svnroot/paradiseo@1804 331e1502-861f-0410-8da2-ba01fb791d7f

trunk/paradiseo-mo/src/mo.h

@@ -158,6 +158,7 @@
 #include <sampling/moMHRndFitnessCloudSampling.h>
 #include <sampling/moMHBestFitnessCloudSampling.h>
 #include <sampling/moNeutralWalkSampling.h>
+#include <sampling/moStatistics.h>
 
 #include <problems/bitString/moBitNeighbor.h>
 #include <problems/eval/moOneMaxIncrEval.h>

trunk/paradiseo-mo/src/sampling/moStatistics.h

@@ -0,0 +1,215 @@
+/*
+  <moStatistics.h>
+  Copyright (C) DOLPHIN Project-Team, INRIA Lille - Nord Europe, 2006-2010
+
+  Sebastien Verel, Arnaud Liefooghe, Jeremie Humeau
+
+  This software is governed by the CeCILL license under French law and
+  abiding by the rules of distribution of free software.  You can  use,
+  modify and/ or redistribute the software under the terms of the CeCILL
+  license as circulated by CEA, CNRS and INRIA at the following URL
+  "http://www.cecill.info".
+
+  As a counterpart to the access to the source code and  rights to copy,
+  modify and redistribute granted by the license, users are provided only
+  with a limited warranty  and the software's author,  the holder of the
+  economic rights,  and the successive licensors  have only  limited liability.
+
+  In this respect, the user's attention is drawn to the risks associated
+  with loading,  using,  modifying and/or developing or reproducing the
+  software by the user in light of its specific status of free software,
+  that may mean  that it is complicated to manipulate,  and  that  also
+  therefore means  that it is reserved for developers  and  experienced
+  professionals having in-depth computer knowledge. Users are therefore
+  encouraged to load and test the software's suitability as regards their
+  requirements in conditions enabling the security of their systems and/or
+  data to be ensured and,  more generally, to use and operate it in the
+  same conditions as regards security.
+  The fact that you are presently reading this means that you have had
+  knowledge of the CeCILL license and that you accept its terms.
+
+  ParadisEO WebSite : http://paradiseo.gforge.inria.fr
+  Contact: paradiseo-help@lists.gforge.inria.fr
+*/
+
+#ifndef moStatistics_h
+#define moStatistics_h
+
+#include <vector>
+#include <utils/eoDistance.h>
+
+/**
+ * Tools to compute some basic statistics
+ *   
+ *   Remember it is better to use some statistic tool like R, etc.
+ *   But it could be usefull to have here in paradisEO
+ */
+class moStatistics 
+{
+public:
+  /**
+   * Default Constructor
+   */
+  moStatistics() { }
+
+  /**
+   * To compute min, max , average and standard deviation of a vector of double
+   *
+   * @param data vector of double 
+   * @param min to compute
+   * @param max to compute
+   * @param avg average to compute
+   * @param std standard deviation to compute
+   */
+  void basic(const vector<double> & data, 
+	double & min, double & max, double & avg, double & std) {
+    
+    if (data.size() == 0) {
+      min = 0.0;
+      max = 0.0;
+      avg = 0.0;
+      std = 0.0;
+    } else {
+      unsigned int n = data.size();
+
+      min = data[0];
+      max = data[0];
+      avg = 0.0;
+      std = 0.0;   
+
+      double d;
+      for(unsigned int i = 0; i < n; i++) {
+	d = data[i];
+	if (d < min)
+	  min = d;
+	if (max < d)
+	  max = d;
+	avg += d;
+	std += d * d;
+      }
+
+      avg /= n;
+
+      std = std / n - avg * avg ;
+      if (std > 0)
+	std = sqrt(std);
+    }
+  }
+
+  /**
+   * To compute the distance between solutions
+   *
+   * @param data vector of solutions
+   * @param distance method to compute the distance
+   * @param matrix matrix of distances between solutions
+   */
+  template <class EOT>
+  void distances(const vector<EOT> & data, eoDistance<EOT> & distance,
+		 vector< vector<double> > & matrix) {
+    if (data.size() == 0) {
+      matrix.resize(0);
+    } else {
+      unsigned int n = data.size();
+
+      matrix.resize(n);
+      for(unsigned i = 0; i < n; i++)
+	matrix[i].resize(n);
+
+      unsigned j;
+      for(unsigned i = 0; i < n; i++) {
+	matrix[i][i] = 0.0;
+	for(j = 0; j < i; j++) {
+	  matrix[i][j] = distance(data[i], data[j]);
+	  matrix[j][i] = matrix[i][j];
+	}
+      }
+    }
+  }
+
+  /**
+   * To compute the autocorrelation and partial autocorrelation
+   *
+   * @param data vector of double
+   * @param nbS number of correlation coefficients
+   * @param rho autocorrelation coefficients
+   * @param phi partial autocorrelation coefficients
+   */
+  void autocorrelation(const vector<double> & data, unsigned int nbS,
+		       vector<double> & rho, vector<double> & phi) {
+    if (data.size() == 0) {
+      rho.resize(0);
+      phi.resize(0);
+    } else {
+      unsigned int n = data.size();
+
+      double cov[nbS+1];
+      double m[nbS+1];
+      double sig[nbS+1];
+
+      rho.resize(nbS+1);
+      phi.resize(nbS+1);
+      rho[0] = 1.0;
+      phi[0] = 1.0; // ?
+      
+      unsigned s, k;
+
+      for(s = 0; s <= nbS; s++) {
+	cov[s] = 0;
+	m[s]   = 0;
+	sig[s] = 0;
+      }
+
+      double m0, s0;
+      unsigned j;
+
+      k = 0;
+      s = nbS;
+      while (s > 0) {
+	while (k + s < n) {
+	  for(j = 0; j <= s; j++) {
+	    m[j]   += data[k+j];
+	    sig[j] += data[k+j] * data[k+j];
+	    cov[j] += data[k] * data[k+j];
+	  }
+	  k++;
+	}
+      
+	m[s]  /= n - s;
+	sig[s] = sig[s] / (n - s) - m[s] * m[s];
+	if (sig[s] <= 0)
+	  sig[s] = 0;
+	else
+	  sig[s] = sqrt(sig[s]);
+	m0 = m[0] / (n - s);
+	s0 = sqrt(sig[0] / (n - s) - m0 * m0);
+	cov[s] = cov[s] / (n - s) - (m[0] / (n - s)) * m[s];
+	rho[s] = cov[s] / (sig[s] * s0);
+	s--;
+      }
+
+      double phi2[nbS+1][nbS+1];
+      double tmp1, tmp2;
+
+      phi2[1][1] = rho[1];
+      for(k = 2; k <= nbS; k++) {
+	tmp1 = 0;
+	tmp2 = 0;
+	for(j = 1; j < k; j++) {
+	  tmp1 += phi2[k-1][j] * rho[k-j];
+	  tmp2 += phi2[k-1][j] * rho[j];
+	}
+	phi2[k][k] = (rho[k] - tmp1) / (1 - tmp2);
+	for(j = 1; j < k; j++) 
+	  phi2[k][j] = phi2[k-1][j] - phi2[k][k] * phi2[k-1][k-j];
+      }
+      
+      for(j = 1; j <= nbS; j++)
+	phi[j] = phi2[j][j];
+      
+    }
+  }
+
+};
+
+
+#endif

trunk/paradiseo-mo/tutorial/Lesson6/adaptiveWalks.cpp

@@ -193,6 +193,7 @@ void main_function(int argc, char **argv)
 
   std::cout << "Last values:" << std::endl;
   std::cout << "Length  " << lengthValues[lengthValues.size() - 1] << std::endl;
+
 }
 
 // A main that catches the exceptions

trunk/paradiseo-mo/tutorial/Lesson6/autocorrelation.cpp

@@ -40,6 +40,10 @@ using namespace std;
 // the sampling class
 #include <sampling/moAutocorrelationSampling.h>
 
+//-----------------------------------------------------------------------------
+// the statistics class
+#include <sampling/moStatistics.h>
+
 // Declaration of types
 //-----------------------------------------------------------------------------
 // Indi is the typedef of the solution type like in paradisEO-eo
@@ -195,6 +199,16 @@ void main_function(int argc, char **argv)
 
   std::cout << "Last values:" << std::endl;
   std::cout << "Fitness  " << fitnessValues[fitnessValues.size() - 1] << std::endl;
+
+  // more basic statistics on the distribution:
+  moStatistics statistics;
+
+  vector<double> rho, phi;
+
+  statistics.autocorrelation(fitnessValues, 10, rho, phi);
+
+  for(unsigned s = 0; s < rho.size(); s++)
+    std::cout << s << " " << "rho=" << rho[s] << ", phi=" << phi[s] << std::endl; 
 }
 
 // A main that catches the exceptions

trunk/paradiseo-mo/tutorial/Lesson6/densityOfStates.cpp

@@ -34,6 +34,10 @@ using namespace std;
 // the sampling class
 #include <sampling/moDensityOfStatesSampling.h>
 
+//-----------------------------------------------------------------------------
+// the statistics class
+#include <sampling/moStatistics.h>
+
 // Declaration of types
 //-----------------------------------------------------------------------------
 // Indi is the typedef of the solution type like in paradisEO-eo
@@ -165,6 +169,14 @@ void main_function(int argc, char **argv)
 
   std::cout << "Last values:" << std::endl;
   std::cout << "Fitness  " << fitnessValues[fitnessValues.size() - 1] << std::endl;
+
+  // more basic statistics on the distribution:
+  double min, max, avg, std;
+  
+  moStatistics statistics;
+
+  statistics.basic(fitnessValues, min, max, avg, std);
+  std::cout << "min=" << min << ", max=" << max << ", average=" << avg << ", std dev=" << std << std::endl; 
 }
 
 // A main that catches the exceptions

trunk/paradiseo-mo/tutorial/Lesson6/neutralWalk.cpp

@@ -40,6 +40,10 @@ using namespace std;
 #include <utils/eoDistance.h>
 #include <sampling/moNeutralWalkSampling.h>
 
+//-----------------------------------------------------------------------------
+// the statistics class
+#include <sampling/moStatistics.h>
+
 // Declaration of types
 //-----------------------------------------------------------------------------
 // Indi is the typedef of the solution type like in paradisEO-eo
@@ -209,6 +213,27 @@ void main_function(int argc, char **argv)
 
   // export only the solution into file
   sampling.fileExport(0, str_out + "_sol");
+
+  // more basic statistics on the distribution:
+  moStatistics statistics;
+
+  vector< vector<double> > dist;
+  vector<double> v;
+
+  statistics.distances(solutions, distance, dist);
+
+  for(unsigned i = 0; i < dist.size(); i++) {
+    for(unsigned j = 0; j < dist.size(); j++) {
+      std::cout << dist[i][j] << " " ; 
+      if (j < i)
+	v.push_back(dist[i][j]);
+    }
+    std::cout << std::endl; 
+  }
+
+  double min, max, avg, std;
+  statistics.basic(v, min, max, avg, std);
+  std::cout << "min=" << min << ", max=" << max << ", average=" << avg << ", std dev=" << std << std::endl; 
 }
 
 // A main that catches the exceptions