optimized eoProportional and added universalselect

eo/src/eoProportionalSelect.h

@@ -32,11 +32,13 @@
 #include <utils/eoRNG.h>
 #include <utils/selectors.h>
 #include <eoSelectOne.h>
-
+#include <eoPop.h>
 //-----------------------------------------------------------------------------
 /** eoProportionalSelect: select an individual proportional to her stored fitness
     value 
-
+    
+    Changed the algorithm to make use of a cumulative array of fitness scores, 
+    This changes the algorithm from O(n) per call to  O(log n) per call. (MK)
 */
 //-----------------------------------------------------------------------------
 
@@ -44,8 +46,7 @@ template <class EOT> class eoProportionalSelect: public eoSelectOne<EOT>
 {
 public:
   /// Sanity check
-  eoProportionalSelect(const eoPop<EOT>& pop = eoPop<EOT>()): 
+  eoProportionalSelect(const eoPop<EOT>& pop = eoPop<EOT>()) 
-    total((pop.size() == 0) ? -1.0 : sum_fitness(pop))
   {
     if (minimizing_fitness<EOT>())
       throw std::logic_error("eoProportionalSelect: minimizing fitness");
@@ -53,18 +54,32 @@ public:
 
   void setup(const eoPop<EOT>& _pop)
   {
-    total = sum_fitness(_pop);
+      if (_pop.size() == 0) return;
+      
+      cumulative.resize(_pop.size());
+      cumulative[0] = _pop[0].fitness();
+
+      for (unsigned i = 1; i < _pop.size(); ++i) 
+      {
+	  cumulative[i] = _pop[i].fitness() + cumulative[i-1];
+      }
   }
     
-  /** do the selection, call roulette_wheel. 
+  /** do the selection,  
    */
   const EOT& operator()(const eoPop<EOT>& _pop) 
   {
-    return roulette_wheel(_pop, total) ;
+      if (cumulative.size() == 0) setup(_pop);
+      
+      double fortune = rng.uniform() * cumulative.back(); 
+      typename FitVec::iterator result = std::upper_bound(cumulative.begin(), cumulative.end(), fortune);
+      return _pop[result - cumulative.begin()];
   }
 
 private :
-  typename EOT::Fitness total;
+
+  typedef std::vector<typename EOT::Fitness> FitVec;
+  FitVec cumulative;
 };
 
 #endif

eo/src/eoStochasticUniversalSelect.h

@@ -0,0 +1,102 @@
+// -*- mode: c++; c-indent-level: 4; c++-member-init-indent: 8; comment-column: 35; -*-
+
+//-----------------------------------------------------------------------------
+// eoStochasticUniversalSelect.h
+// (c) Maarten Keijzer 2003
+/* 
+    This library is free software; you can redistribute it and/or
+    modify it under the terms of the GNU Lesser General Public
+    License as published by the Free Software Foundation; either
+    version 2 of the License, or (at your option) any later version.
+
+    This library is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+    Lesser General Public License for more details.
+
+    You should have received a copy of the GNU Lesser General Public
+    License along with this library; if not, write to the Free Software
+    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+
+    Contact: todos@geneura.ugr.es, http://geneura.ugr.es
+             Marc.Schoenauer@polytechnique.fr
+             mkeijzer@cs.vu.nl
+ */
+//-----------------------------------------------------------------------------
+
+#ifndef eoStochasticUniversalSelect_h
+#define eoStochasticUniversalSelect_h
+
+//-----------------------------------------------------------------------------
+
+#include <utils/eoRNG.h>
+#include <eoSelectOne.h>
+
+//-----------------------------------------------------------------------------
+/** eoStochasticUniversalSelect: select an individual proportional to her stored fitness
+    value, but in contrast with eoStochasticUniversalSelect, get rid of most finite sampling effects
+    by doing all selections in one go, using a single random number.
+*/
+//-----------------------------------------------------------------------------
+
+template <class EOT> class eoStochasticUniversalSelect: public eoSelectOne<EOT> 
+{
+public:
+  /// Sanity check
+  eoStochasticUniversalSelect(const eoPop<EOT>& pop = eoPop<EOT>()) 
+  {
+    if (minimizing_fitness<EOT>())
+      throw std::logic_error("eoStochasticUniversalSelect: minimizing fitness");
+  }
+
+  void setup(const eoPop<EOT>& _pop)
+  {
+      if (_pop.size() == 0) return;
+      
+      std::vector<typename EOT::Fitness> cumulative(_pop.size());
+      
+      cumulative[0] = _pop[0].fitness();
+      for (unsigned i = 1; i < _pop.size(); ++i) 
+      {
+	  cumulative[i] = _pop[i].fitness() + cumulative[i-1];
+      }
+      
+      indices.reserve(_pop.size());
+      indices.resize(0);
+
+      double fortune = rng.uniform() * cumulative.back();
+      double step = cumulative.back() / double(_pop.size());
+      
+      unsigned i = std::upper_bound(cumulative.begin(), cumulative.end(), fortune) - cumulative.begin();
+      
+      while (indices.size() < _pop.size()) {
+	    
+	  while (cumulative[i] < fortune) {i++;} // linear search is good enough as we average one step each time
+	  
+	  indices.push_back(i);
+	  fortune += step;
+	  if (fortune >= cumulative.back()) { // start at the beginning
+	      fortune -= cumulative.back();
+	      i = 0;
+	  }
+      }
+  }
+    
+  /** do the selection,  
+   */
+  const EOT& operator()(const eoPop<EOT>& _pop) 
+  {
+      if (indices.empty()) setup(_pop);
+    
+      unsigned index = indices.back();
+      indices.pop_back();
+      return _pop[index];
+  }
+
+private :
+
+  typedef std::vector<unsigned> IndexVec;
+  IndexVec indices;
+};
+
+#endif