main()
{
Pop pop;
EORandom<float> random;
Chrom chrom(NUM_GENES, BITSxGEN);
for (unsigned i = 0; i < POP_SIZE; i++)
{
random(chrom);
chrom.fitness(i);
pop.push_back(chrom);
}
Bin fitness;
EOEvalAll<Chrom> eval(fitness);
EOStat<Chrom> graphEval( eval, true ); // True is for verbose
EORank<Chrom> transform(1);
// EOMutation<Chrom> mutation(0.05);
// transform.addOp((EOOp<Chrom>*)&mutation);
NxOver<Chrom> nxover(2);
transform.addOp(&nxover);
EOGenTerm<Chrom> term(100);
EOAgeGA<Chrom> agega(graphEval, transform, term);
try
{
agega(pop);
}
catch(UException& e)
{
cout << e.what() << endl;
}
cout << "the best: " << pop.front() << endl;
return 0;
}
main( int argc, char** argv) {
unsigned numGenes = 2, // Number of genes in the chromosome
numGenerations, // Number of generations
popSize; // Population size
float xOverPr, // Crossover rate, from 0 to 1
mutPr; // Mutation rate, from 0 to 1
bool verbose; // true if you want many things printed
// Obtain command line parameters or default values
getParams( argc, argv, verbose, popSize, numGenerations,
xOverPr, mutPr );
// Create an initial population and the population-level operators
Pop pop;
unsigned i, j;
Uniform<float> u( 0, 1);
for ( j = 0; j < popSize; j ++ ) {
Chrom* aChrom = new Chrom;
for ( i = 0; i < numGenes; i ++ ) {
aChrom->push_back( u() );
}
pop.push_back( *aChrom );
if ( verbose ) {
copy( aChrom->begin(), aChrom->end(), ostream_iterator<int>( cout, " ") );
cout << endl;
}
delete aChrom;
}
// Create an instance of the evaluation function object
deJongF2float thisEvalFunc;
// From that, create an object that evaluates all the population
EOEvalAll<Chrom> eval( thisEvalFunc) ;
// Then, an object that select who's going to reproduce
EOLottery<Chrom> select;
// Another object that mates them
EORandomBreed<Chrom> breed;
// And another that weeds out the worst
EOElimAll<Chrom> replace;
// This one check if the algorithm has finishd or not
EOGenTerm<Chrom> term( numGenerations );
// Then the EO-level operators
// This is a mutation-like operator
EOFloatArrayRndMutate<Chrom> rndMut( 0.1, mutPr );
// And this is the classical 2-point crossover
EOXOver2<Chrom> xOver(xOverPr);
// Add operators to breeder
breed.addOp( &rndMut );
breed.addOp( &xOver );
// Then, the algorithm itself is instantiated
EOEasyGA<Chrom> ga( eval, select, breed, term, replace, verbose );
// Now, apply the algorithm to the population
eval( pop );
ga( pop );
// Sort Population. First is the best
sort( pop.begin(), pop.end(), SortEO<Chrom>() );
// And print results!
cout << "Best fitness ....... " << pop[0].fitness() << endl
<< "Value............... " << pop[0] << endl;
return 0;
}
void operator()()
{
communicator world;
ostringstream ss;
ss << "trace." << world.rank() << ".algo.txt";
ofstream of(ss.str().c_str());
for (int g = 0; g >= 0/* < 10*/; ++g)
{
// of << "."; of.flush();
// of << _pop.size() << " "; of.flush();
// Selection
for (int i = 0; i < world.size(); ++i)
{
if (i == world.rank()) continue;
// int chunk = 2;
// int size = _pop.size() > chunk ? chunk : _pop.size();
int size = _pop.size();
if (size)
{
Pop<EOT> newpop;
// copy_n(_pop.begin(), size, back_inserter(newpop));
for (auto& ind : _pop)
{
newpop.push_back(ind);
}
of << _pop.size() << " "; of.flush();
// _pop.erase(_pop.begin(), _pop.begin() + size);
_pop.clear();
of << _pop.size() << " "; of.flush();
m_em.lock();
vec_em[i].push( newpop );
m_em.unlock();
}
}
// !Selection
for (int i = 0; i < world.size(); ++i)
{
if (i == world.rank()) continue;
m_imm.lock();
while (!vec_imm[i].empty())
{
of << "="; of.flush();
auto& newpop = vec_imm[i].front();
// copy(newpop.begin(), newpop.end(), back_inserter(_pop));
for (auto& ind : newpop)
{
_pop.push_back(ind);
}
vec_imm[i].pop();
}
m_imm.unlock();
}
// std::this_thread::sleep_for(std::chrono::microseconds(100000));
pt_cv->notify_one();
}
}