HillClimbing::HillClimbing()
: OptimisationAlgorithm() {
bestFoundParameter_ = 2.0 * arma::ones<arma::vec>(3);
setInitialisingFunctions({{[this](
const arma::uword numberOfDimensions_,
const arma::mat& initialParameters_) {
if (getMinimalStepSize() > getMaximalStepSize()) {
throw std::logic_error("HillClimbing.initialisingFunctions: The maximal step size must be must be greater than or equal to the minimal one.");
}
return initialParameters_;
},
"Step size validation"}});
setNextParametersFunctions({{[this](
const arma::uword numberOfDimensions_,
const arma::mat& parameters_,
const arma::rowvec& objectiveValues_,
const arma::rowvec& differences_) {
return randomNeighbour(getBestFoundParameter(), getMinimalStepSize(), getMaximalStepSize());
},
"Return a random neighbour of the best parameter within the step sizes"}});
setMinimalStepSize(0.0);
setMaximalStepSize(0.1);
}
SimulatedAnnealing::SimulatedAnnealing()
: OptimisationAlgorithm() {
setInitialisingFunctions({{
[this](
const arma::uword numberOfDimensions_,
const arma::mat& initialParameters_) {
if (minimalStepSize_ > maximalStepSize_) {
throw std::logic_error("SimulatedAnnealing.initialisingFunctions: The maximal step size must be must be greater than or equal to the minimal one.");
}
return initialParameters_;
},
"Step size validation"
}
});
setNextParametersFunctions({{
[this](
const arma::uword numberOfDimensions_,
const arma::mat& parameters_,
const arma::rowvec& objectiveValues_,
const arma::rowvec& differences_) {
if (differences_(0) < 0) {
state_ = getBestFoundParameter();
} else {
double durationEnergy = getUsedDuration() / getMaximalDuration();
double numberOfIterationsEnergy = getUsedNumberOfIterations() / getMaximalNumberOfIterations();
double energy;
if (durationEnergy > numberOfIterationsEnergy) {
energy = std::exp(-(objectiveValues_(0) - getBestFoundObjectiveValue()) / (1.0 - durationEnergy));
} else {
energy = std::exp(-(objectiveValues_(0) - getBestFoundObjectiveValue()) / (1.0 - numberOfIterationsEnergy));
}
if (std::bernoulli_distribution(energy)(Rng::generator_)) {
state_ = parameters_.col(0);
}
}
return randomNeighbour(state_, minimalStepSize_, maximalStepSize_);
},
"Simulated annealing"
}
});
setMinimalStepSize(0);
setMaximalStepSize(0.1);
}
SimulatedAnnealing::SimulatedAnnealing(
const std::shared_ptr<OptimisationProblem> optimisationProblem)
: TrajectoryBasedOptimisationAlgorithm(optimisationProblem) {
setMaximalStepSize((getUpperBounds() - getLowerBounds()) * 0.1);
}
HillClimbing<T>::HillClimbing(
const std::shared_ptr<OptimisationProblem<T>> optimisationProblem) noexcept
: TrajectoryBasedOptimisationAlgorithm<T>(optimisationProblem) {
setMaximalStepSize(arma::norm(this->getLowerBounds(), this->getUpperBounds()) / static_cast<T>(10.0L));
}
