AlgGeneticAnnealing.cpp

#include "ECF_base.h"
#include "ECF_macro.h"
#include "AlgGeneticAnnealing.h"
 
 
GeneticAnnealing::GeneticAnnealing()
{
    // define algorithm name
    name_ = "GeneticAnnealing";
 
    selBestOp_ = static_cast<SelBestOpP> (new SelBestOp);
}
 
 
void GeneticAnnealing::registerParameters(StateP state)
{
    registerParameter(state, "energybank", (voidP) new double(200), ECF::DOUBLE,
        "total starting energy bank (fitness dependant)");
    registerParameter(state, "coolingfactor", (voidP) new double (0.7), ECF::DOUBLE,
        "simulated annealing cooling factor"); 
    registerParameter(state, "elitism", (voidP) new uint (0), ECF::UINT,
        "is the best individual preserved"); 
}
 
 
bool GeneticAnnealing::initialize(StateP state)
{
    // read parameter values 
    voidP p = getParameterValue(state, "energybank");
    this->energyBank_ = *((double*) p.get());
 
    if(energyBank_ <= 0) {
        ECF_LOG(state, 1, "Error: GeneticAnnealing algorithm parameter 'energybank' must be greater than zero!");
        throw "";
    }
 
    p = getParameterValue(state, "coolingfactor");
    this->coolingF_ = *((double*) p.get());
 
    if ((coolingF_ < 0)||(coolingF_>=1)){
        ECF_LOG(state, 1, "Error: GeneticAnnealing algorithm parameter 'coolingfactor' must be in <0, 1]!");
        throw "";
    }
 
    p = getParameterValue(state, "elitism");
    elitism_ = *((uint*) p.get()) ? true : false;
 
    isFirstGeneration_ = true;
 
    selBestOp_->initialize(state);
 
    return true;
}
 
 
bool GeneticAnnealing::advanceGeneration(StateP state, DemeP deme)
{
    uint N = (uint) deme->size();
    uint i;
    IndividualP mutant, ind;
 
    IndividualP best = selBestOp_->select(*deme);
 
    if (isFirstGeneration_){
        Th = (double *) malloc (N*sizeof(double));
        for (i = 0; i < N; i++){
            ind = deme->at(i);
            Th[i] = ind->fitness->getValue() + energyBank_/N;
        } // init of threshold
 
        energyBank_ = 0;
        isFirstGeneration_ = false;
    }
 
    for (i = 0; i < N; i++){
        ind = deme->at(i);
 
        if(elitism_ && ind == best)
            continue;
 
        mutant = (IndividualP) ind->copy();
        mutate (mutant); // create mutant
        evaluate(mutant);
 
        double eMutant = mutant->fitness->getValue();
        
        if (eMutant <= Th[i]){
            energyBank_ = energyBank_ + Th[i] - eMutant;
            Th[i] = eMutant;
            replaceWith(ind, mutant); // replace individual with mutant
        }
    }
 
    dE_ = energyBank_ * coolingF_ / N;
    ECF_LOG(state, 4, "GeneticAnnealing: energy bank=" + dbl2str(energyBank_) + ", dE_=" + dbl2str(dE_));
 
    energyBank_ = 0;
    for (i = 0; i < N; i++)
        Th[i] += dE_;
 
    return true;
}