C++ Population::evaluate方法代码示例

/*
 * single_run.cpp
* it is basically the independet thread that can be launched
* independently
 *  Created on: Mar 2, 2011
 *      Author: cnua
 */
int single_run (int run_id, int cur_run, int* SEED, string OUTPUT_STRING, \
		RunParameters* p_parameters,\
		ProblemDefinition* p_problem, \
		Reporter pop_reporter,\
		time_t start, time_t finish, double delta_t)
{
	// create directory run_k
	char num_field[10];
	sprintf(num_field, "%d", cur_run+1);
	string num_field_s = num_field;
	string DIR_RUN_K =  OUTPUT_STRING + "/run_" + num_field;
	mkdir(DIR_RUN_K.c_str(), S_IRWXU | S_IRWXG | S_IROTH | S_IXOTH );

	// seed the random generator
	srand(SEED[cur_run]);

	// start time
	time(&start);
	int elapsed_time;

	//---------------------------------------------------------------------
	// CREATE THE POPULATION (constructor called)
	//---------------------------------------------------------------------
	Variable* Z;    //to be included in ProblemDefinition...
	p_problem->initialise_variables(&Z, p_parameters->max_n_periods);

	// do you really need new? no, but don't touch it for now
	Population* P = new Population(p_parameters, p_problem);
	if (!P) {
		cerr << "\nmain : Error creating population!!!\n";
		exit(-1);
	}

	//as it's hard to pass Pop as a parameter to fdf_c__ through fortran functions, treat it as a global variable
	Pop = P;
	cout << "\nP = " << P;
	cout << "\nPop = " << Pop;

	///// INITIAL GENERATION (0) ///////////////////////////////////////

	// trees before parameters insertion
			cout << "\nInitialization of the population (generation 0)\n";
			P->print_population_without_parameters(0);


			/// split the data set in tuning set (data_tuning) and validation set (data_validation). See SPLIT and VALIDATING_LINES in input file
			//this function will also allow to increase the number of fitness cases during the run...
			P->split_data(p_parameters, p_problem, 0,1);
			// here or before parallel section begins?
			// Depends if you want to change fitness cases during the evolution...


			/////////// OTHER GENERATIONS ///////////////////////////////////////
			int check_end = 0;
			int last_gen=0;
			for (int i=0; i<p_parameters->G+1; i++) {   // generations

				if (i) {   //skip generation 0
					// split the data for the current generation (this function will also allow to increase the number of fitness cases during the run...)
					//P->split_data(i,G,split); // not used...
				/*
					if ((i%6)==0) {   //6
						// KILLING and FILLING
						P->kill_and_fill(&problem);
						//cin.get();
					}
					else
				//*/
					// GENETIC OPERATORS: sorting, reproduction, crossover, mutation, pruning
					P->new_spawn(*p_parameters, *p_problem, p_parameters->nfitcases,i);

					// print population WITHOUT parameters after genetic operations
					//if (COMMENT) {
					//	printf("\n\n***********Generation %d after genetic operations (not sorted, new trees marked with f9.999999E+99)************\n", i-1);
					//	P->print_population_without_parameters(i-1);
					//	printf("\n***********************************************************************\n");
					//}
				}

				// evaluate fitness function (in structural GP parameters are added and tuned first, then the evaluation is performed)
				P->evaluate(i,p_parameters->G);

				//----------------------------------------------------------------------------------
				// extfitness - keeping the individual with least fitness value -------
				// sort according to fitness (error)
				//P->sort(i,tree_comp_fitness);

				///printf("\n\n***********Generation %d after genetic operations (not sorted, new trees marked with f9.999999E+99)************\n", i-1);
				//P->print_population_without_parameters(i-1);
				//printf("\n***********************************************************************\n");

				// update the best individual - structure and complete tree (for PARAMETER INHERITANCE)
				//P->update_ext_archive();
//.........这里部分代码省略.........