selectio.h

改算法基于遗传算法

//                          Time-stamp: <2007-07-12 11:55:58 handa>
// studied  by T.Noda        99/08/05  Ver.0.1
// modified by H.Handa       99/08/16  Ver.0.4
// gcc ___.cc -lstdc++

#ifndef SELECTIO_H
#define SELECTIO_H

#include "chromoso.h"
#include <vector>
#include <list>

using namespace std;


template <class T> class selection {
protected:
  random_number& rn;
public:
  selection(random_number& r): rn(r) { }  

  virtual void mating(list<chromosome<T> *>& pop,
		      vector<chromosome<T> *>& pool,
		      int flag)
    {

      int i,j;
 
      list<chromoIntP>::iterator il;
      double sum_fit=0.0;
      double bias = 0.0;
      for(il = pop.begin() ;il != pop.end();il++) {
	sum_fit += (*il)->fitness(0);
	bias = (bias > (*il)->fitness(0)) ? (*il)->fitness(0) : bias;
      }
      sum_fit -= bias*pop.size();
      for(i = 0 ;i < pool.size();i++) {
	double dtmp = rn.urand(sum_fit);
	for(il = pop.begin() ;il != pop.end();il++) {
	  dtmp -= (*il)->fitness(0) - bias;
	  if(dtmp <= 0.0) break;
	}  
	if ( il == pop.end()) {
	  cerr << "error may be occured in  population::" <<
	    "roulette_preselect()" << endl;
	  j = pop.size() -1;
	}
	pool[i] = (*il);
	pool[i]->setFlag(flag);
      }


    }
  // this class indicate roulette wheel selection, which implies
  // that our fitness is uni-objective and maxmize one.
  virtual void postprocessing(list<chromosome<T> *>& pool,
			      list<chromosome<T> *>& pop, int flag)
    {
      list<chromoIntP>::iterator il;
      int flag2 = 1-flag;

      for(il = pop.begin(); il != pop.end(); il++) {
	while ((*il)->isFlag() == flag2) {
	  il = pop.erase(il);
	}
      }
      for(il = pool.begin(); il != pool.end();il++) {
	(*il)->setFlag(flag);
	(*il)->fitness() = (*il)->objfunc();
	pop.push_front(*il);
      }


  }

  virtual selection<T>& operator=(const selection<T>& arg) {
    if(this == &arg) return *this;
    rn = arg.rn;     // random
  }

};


#endif