void reproduce( population_type& pop , fitness_type& fitness )
{
GPCXX_ASSERT( pop.size() == fitness.size() );
GPCXX_ASSERT( m_rates.size() == m_operators.size() );
GPCXX_ASSERT( m_operators.size() > 0 );
std::vector< size_t > indices;
sort_indices( fitness , indices );
population_type new_pop;
new_pop.reserve( pop.size() );
// elite
for( size_t i=0 ; i<m_number_elite ; ++i )
{
index_vector elite_in_indices;
index_vector elite_out_indices;
size_t index = indices[i] ;
elite_in_indices.push_back( index );
elite_out_indices.push_back( new_pop.size() );
new_pop.push_back( pop[ index ] );
m_observer( -1 , elite_in_indices , elite_out_indices );
}
size_t n = pop.size();
std::discrete_distribution< int > dist( m_rates.begin() , m_rates.end() );
while( new_pop.size() < n )
{
int choice = dist( m_rng );
auto& op = m_operators[ choice ];
auto selection = op.selection( pop , fitness );
index_vector in;
for( auto s : selection ) in.push_back( s - pop.begin() );
auto trees = op.operation( selection );
index_vector out;
for( auto iter = trees.begin() ; ( iter != trees.end() ) && ( new_pop.size() < n ) ; ++iter )
{
out.push_back( new_pop.size() );
m_final_transform( *iter );
new_pop.push_back( std::move( *iter ) );
}
m_observer( choice , in , out );
}
pop = std::move( new_pop );
}
// setup runtime and population for the genetic algorithm
genetic_algorithm( const std::size_t runtime_in_second_=3600, const std::size_t population_per_generation_ = 256 ): population_per_generation(population_per_generation_)
{
current_population.reserve( population_per_generation );
selected_population_for_xover_father.reserve( population_per_generation );
selected_population_for_xover_mother.reserve( population_per_generation );
//Should all these managers move to private member list????
//setup timer
auto& t_manager = feng::singleton<time_manager>::instance();
t_manager(runtime_in_second_);
//setup selection manager
auto& xs_manager = feng::singleton<xover_selection_manager>::instance();
xs_manager.initialize( population_per_generation );
//setup mutation manager
auto& m_manager = feng::singleton<mutation_manager>::instance();
m_manager.initialize( population_per_generation );
}
Problem_Representation_Type const
operator()( Chromosome_Args ... ch_args )
{
//initialize first generation randomly
for ( std::size_t i = 0; i != population_per_generation; ++i )
current_population.push_back( chromosome_type( ch_args... ) );
//std::cerr << "\ninitialized " << population_per_generation << " chromosomes.\n";
std::size_t debug_counter = 0;
best_fitness = std::numeric_limits<Fitness_Type>::max();
srand ( unsigned ( time (NULL) ) ); //for random_shuffle
Fitness_Type total_fit;
Problem_Representation_Type pr_0;
Problem_Representation_Type pr_1;
Problem_Representation_Type pr_2;
Problem_Representation_Type pr_3;
std::ofstream elite_out( "elite.dat" );
std::ofstream elite_fit( "elite_fit.dat");
for (;;)
{
//evaluate
//feng::for_each( current_population.begin(), current_population.end(), [](chromosome_type& chrom) { Evaluation_Method()(Chromosome_Problem_Translation_Method()(chrom)); });
//std::cerr << "\nevaluating........";
#if 0
for ( auto& ch : current_population )
if ( ch.modification_after_last_evaluation_flag )
{
Chromosome_Problem_Translation_Method()( *(ch.chrom), pr );
ch.fit = Evaluation_Method()(pr);
//Evaluation_Method()(Chromosome_Problem_Translation_Method()(ch));
ch.modification_after_last_evaluation_flag = false;
}
#endif
#if 1
std::thread t0( parallel_evaluator(), current_population.begin(), current_population.begin()+(current_population.size()/4), &pr_0 );
std::thread t1( parallel_evaluator(), current_population.begin()+(current_population.size()/4), current_population.begin()+(current_population.size()/2), &pr_1 );
std::thread t2( parallel_evaluator(), current_population.begin()+(current_population.size()/2), current_population.begin()+(current_population.size()*3/4), &pr_2 );
std::thread t3( parallel_evaluator(), current_population.begin()+(current_population.size()*3/4), current_population.end(), &pr_3 );
//parallel_evaluator()( current_population.begin()+(current_population.size()*3/4), current_population.end(), &pr_3 );
t0.join();
t1.join();
t2.join();
t3.join();
#endif
//using nth_element?
//mutate duplicated chromosome?
std::sort( current_population.begin(), current_population.end() );
auto const elite_one = *(current_population.begin());
Chromosome_Problem_Translation_Method()( *(elite_one.chrom), pr );
//keep the best individual of the current generation
if ( elite_one.fit < best_fitness )
{
debug_counter++;
best_fitness = elite_one.fit;
best_one = pr;
elite_out << best_one;
elite_fit << best_fitness;
if ( debug_counter & 0xf )
{
elite_out << "\n";
elite_fit << "\n";
}
else
{
elite_out << std::endl;
elite_fit << std::endl;
}
}
current_population.resize( std::distance( current_population.begin(), std::unique( current_population.begin(), current_population.end() )) );
if ( current_population.size() < population_per_generation )
{
//generate someone randomly and evaluate
for ( std::size_t i = current_population.size(); i != population_per_generation; ++i )
{
auto ch = chromosome_type( ch_args...);
Chromosome_Problem_Translation_Method()( *(ch.chrom), pr );
ch.fit = Evaluation_Method()(pr);
ch.modification_after_last_evaluation_flag = false;
current_population.push_back( ch );
}
}
else
{
//shuffle the resize
std::random_shuffle( current_population.begin(), current_population.end() );
current_population.resize( population_per_generation );
}
std::sort( current_population.begin(), current_population.end() );
//total_fit = 0;
//for( auto& ch : current_population )
// total_fit += ch.fit;
//std::cout.precision(20);
//std::cout << total_fit << "\n";
//if timeout, return output elite
auto& t_manager = feng::singleton<time_manager>::instance();
if ( t_manager.is_timeout() )
{
std::cerr << "\nthe residual for the best individual is " << best_fitness;
elite_out.close();
elite_fit.close();
return best_one;
//return pr;
}
//std::cerr << "\nElite of " << debug_counter++ << " is \n" << pr;
//std::cerr << "\nElite of " << debug_counter << " is \n" << *(elite_one.chrom);
//select
auto& xpm = feng::singleton<xover_probability_manager<>>::instance();
std::size_t const selection_number = xpm(population_per_generation);
//selected_population_for_xover_father.resize(selection_number);
//selected_population_for_xover_mother.resize(selection_number);
selected_population_for_xover_father.clear();
selected_population_for_xover_mother.clear();
//std::cerr << "\nselecting............." << selection_number;
auto& xs_manager = feng::singleton<xover_selection_manager>::instance();
for ( std::size_t i = 0; i != selection_number; ++i )
{
//selected_population_for_xover_father[i] = current_population[xs_manager()];
//selected_population_for_xover_mother[i] = current_population[xs_manager()];
const std::size_t select1 = xs_manager();
const std::size_t select2 = xs_manager();
//selected_population_for_xover_father.push_back(current_population[xs_manager()]);
//selected_population_for_xover_mother.push_back(current_population[xs_manager()]);
selected_population_for_xover_father.push_back(current_population[select1]);
selected_population_for_xover_mother.push_back(current_population[select2]);
}
//std::cerr << "\nselectedted";
//xover
#if 0
//not working as selected population for xover are pure reference of current populaiton
for ( std::size_t i = 0; i != selection_number; ++i )
feng::for_each( selected_population_for_xover_father[i].begin(), selected_population_for_xover_father[i].end(),
selected_population_for_xover_mother[i].begin(), Chromosome_Xover_Method() );
current_population.reserve( population_per_generation + selection_number + selection_number );
//append newly generated genes into current population
current_population.insert( current_population.begin()+population_per_generation, selected_population_for_xover_father.begin(), selected_population_for_xover_father.end() );
current_population.insert( current_population.begin()+population_per_generation+selection_number, selected_population_for_xover_mother.begin(), selected_population_for_xover_mother.end() );
#endif
//std::cerr << "\nxover.............";
for ( std::size_t i = 0; i != selection_number; ++i )
{
chromosome_type son( ch_args... );
chromosome_type daughter( ch_args... );
feng::for_each( selected_population_for_xover_father[i].begin(), selected_population_for_xover_father[i].end(), selected_population_for_xover_mother[i].begin(), son.begin(), daughter.begin(), Chromosome_Xover_Method() );
current_population.push_back( son );
current_population.push_back( daughter );
}
//mark new generation not evaluated
feng::for_each( current_population.begin()+population_per_generation, current_population.end(), [](chromosome_type& ch) { ch.modification_after_last_evaluation_flag = true; } );
//std::cerr << "\nxovered";
//mutate
//std::cerr << "\nmutating";
auto& mpm = feng::singleton<mutation_probability_manager<>>::instance();
mpm.reset();
for ( auto& chromosome : current_population )
if ( mpm.should_mutate_current_gene() )
{
for ( auto& gene : chromosome )
Chromosome_Mutation_Method()(gene);
//makr muated ones as not evaluated
chromosome.modification_after_last_evaluation_flag = true;
}
//std::cerr << "\nmutated";
//goto evaluate
}
assert( !"Should never reach here!!" );
return Problem_Representation_Type();
}
