void NoisyRprop::init(const ObjectiveFunctionType & objectiveFunction, const SearchPointType& startingPoint,const RealVector& delta0P){
SIZE_CHECK(startingPoint.size()==delta0P.size());
m_best.point=startingPoint;
m_minEpisodeLength = 40;
size_t pc = startingPoint.size();
m_episodeLength.resize(pc);
m_sample.resize(pc);
m_sampleIteration.resize(pc);
m_iteration.resize(pc);
m_toTheRight.resize(pc);
m_position.resize(pc);
m_leftmost.resize(pc);
m_rightmost.resize(pc);
m_leftsum.resize(pc);
m_rightsum.resize(pc);
m_current.resize(pc);
m_numberOfAverages.resize(pc);
m_gradient.resize(pc);
m_delta = delta0P;
std::fill(m_episodeLength.begin(),m_episodeLength.end(),m_minEpisodeLength);
std::fill(m_sample.begin(),m_sample.end(),0);
std::fill(m_sampleIteration.begin(),m_sampleIteration.end(),m_minEpisodeLength / 100);
std::fill(m_iteration.begin(),m_iteration.end(),0);
std::fill(m_position.begin(),m_position.end(),0);
std::fill(m_leftmost.begin(),m_leftmost.end(),0);
std::fill(m_rightmost.begin(),m_rightmost.end(),0);
std::fill(m_toTheRight.begin(),m_toTheRight.end(),0);
m_leftsum.clear();
m_rightsum.clear();
std::fill(m_current.begin(),m_current.end(),0);
std::fill(m_numberOfAverages.begin(),m_numberOfAverages.end(),1);
m_gradient.clear();
}
void RpropMinus::init(
ObjectiveFunctionType & objectiveFunction,
SearchPointType const& startingPoint,
double initDelta
) {
checkFeatures(objectiveFunction);
objectiveFunction.init();
m_parameterSize = startingPoint.size();
m_delta.resize(m_parameterSize);
m_oldDerivative.resize(m_parameterSize);
std::fill(m_delta.begin(),m_delta.end(),initDelta);
m_oldDerivative.clear();
m_best.point = startingPoint;
//evaluate initial point
m_best.value = objectiveFunction.evalDerivative(m_best.point,m_derivative);
}
double eval(const SearchPointType &p) const {
m_evaluationCounter++;
// Setup the parameters
Eigen::VectorXd params( p. size() );
for (int i = 0; i < p.size(); i++) params(i) = p(i);
mPolicy->setParams(params);
mSim->reset();
for (int i = 0; i < 2000; i++) {
mSim->step();
}
double value = mSim->eval()->cost();
LOG(INFO) << "# " << m_evaluationCounter << " : " << params.transpose()
<< " -> " << value;
return value;
}
void AbstractLineSearchOptimizer::init(ObjectiveFunctionType& objectiveFunction, SearchPointType const& startingPoint) {
checkFeatures(objectiveFunction);
objectiveFunction.init();
m_linesearch.init(objectiveFunction);
m_dimension = startingPoint.size();
m_best.point = startingPoint;
m_best.value = objectiveFunction.evalDerivative(m_best.point,m_derivative);
// Get space for the different vectors we store.
m_lastPoint.resize(m_dimension);
m_searchDirection = -m_derivative;
m_initialStepLength = 0.0;//1.0 as step length might be very wrong.
for (size_t i = 0; i < m_derivative.size(); ++i)
m_initialStepLength += std::abs(m_derivative(i));
m_initialStepLength = std::min(1.0, 1.0 / m_initialStepLength);
initModel();
}
void proposeStartingPoint(SearchPointType &x) const {
x.resize(numberOfVariables());
for (unsigned int i = 0; i < x.size(); i++) {
x(i) = shark::Rng::uni(MIN_PARAM, MAX_PARAM);
}
}
void NoisyRprop::init(const ObjectiveFunctionType & objectiveFunction, const SearchPointType& startingPoint,double delta0P){
checkFeatures(objectiveFunction);
RealVector d(startingPoint.size());
std::fill(d.begin(),d.end(),delta0P);
init(objectiveFunction,startingPoint, d);
}