// Use the ratio between the minimum and maximum to determine whether the
// population has converged. This method will not work if the values cross
// zero!
// Note that this is significantly different than the definition (and the
// bug-laden implementation) that was in version of GAlib prior to 2.4.5.
//
// For historical purposes, here is the old definition of this method:
//
// Use the ratio of the population mean divided by the population maximum to
// determine whether the population has converged. If we are maximizing, then
// check to see if the ratio exceeds the convergence. If we are minimizing,
// then check to see if the ratio has dropped below the convergence.
GABoolean
GAGeneticAlgorithm::TerminateUponPopConvergence(GAGeneticAlgorithm & ga){
GABoolean val = gaFalse;
if(ga.statistics().current(GAStatistics::Maximum) == 0) {
return val;
}
float ratio =
ga.statistics().current(GAStatistics::Minimum) /
ga.statistics().current(GAStatistics::Maximum);
if(ga.minimaxi() == GAGeneticAlgorithm::MINIMIZE) {
if(ratio <= ga.pConvergence())
val = gaTrue;
else
val = gaFalse;
}
else {
if(ratio >= ga.pConvergence())
val = gaTrue;
else
val = gaFalse;
}
return val;
}
GABoolean
GAGeneticAlgorithm::TerminateUponPopConvergence(GAGeneticAlgorithm & ga){
if(ga.statistics().current(GAStatistics::Mean) /
ga.statistics().current(GAStatistics::Maximum) < ga.pConvergence())
return gaTrue;
else
return gaFalse;
}
// Send the specified number of individuals from the current population to
// the specified task.
int
SendMigration(int toid, GAGeneticAlgorithm& ga, int count) {
int status = 0;
status = pvm_initsend(PvmDataDefault);
status = pvm_pkint(&count, 1, 1);
for(int i=0; i<count && status>=0; i++)
status = PackIndividual(ga.population().best(i));
status = pvm_send(toid, MSG_INCOMING_MIGRATION);
return status;
}
// Receive a bunch of individuals from a task. To do this, we clone the GA's
// population, stuff the immigrants into the population, trash the worst
// individuals to bring the population size back down to what it was, then
// stick the population back into the GA.
// This implementation is really inefficient, but you get the idea of how
// to do this...
int
RecvMigration(GAGeneticAlgorithm& ga) {
int status = 0;
GAPopulation pop(ga.population());
GAGenome *tmpind = ga.population().individual(0).clone();
int count = 0;
status = pvm_upkint(&count, 1, 1);
for(int i=0; i<count && status>=0; i++) {
status = UnpackIndividual(*tmpind);
pop.add(*tmpind);
}
for(int j=0; j<count; j++)
pop.destroy(); // default behavior is to destroy worst
ga.population(pop);
delete tmpind;
return status;
}
// If we are maximizing, then terminate when the convergence has exceeded the
// specified convergence. If we are minimizing, then terminate when the
// convergence has dropped below the specified convergence.
GABoolean
GAGeneticAlgorithm::TerminateUponConvergence(GAGeneticAlgorithm & ga){
GABoolean val = gaFalse;
if(ga.minimaxi() == GAGeneticAlgorithm::MINIMIZE) {
if(ga.convergence() == 0 || ga.convergence() > ga.pConvergence())
val = gaFalse;
else
val = gaTrue;
}
else {
if(ga.convergence() < ga.pConvergence())
val = gaFalse;
else
val = gaTrue;
}
return val;
}
// Here are a few termination functions that you can use. Terminators return
// gaTrue if the algorithm should finish, gaFalse otherwise.
GABoolean
GAGeneticAlgorithm::TerminateUponGeneration(GAGeneticAlgorithm & ga){
return(ga.generation() < ga.nGenerations() ? gaFalse : gaTrue);
}
void TwitterAnnounce::action(const GAGeneticAlgorithm & ga) {
std::string generation = stringFrom(ga.generation());
// todo get from config
announcer_->announceGeneration("Generation " + generation + " created! #megapolis", "","http://nynex.hydrogenproject.com/rate.php");
}