// -*- mode: c++; c-indent-level: 4; c++-member-init-indent: 8; comment-column: 35; -*- //----------------------------------------------------------------------------- // eoStat.h // (c) Marc Schoenauer, Maarten Keijzer and GeNeura Team, 2000 /* This library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. This library is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details. You should have received a copy of the GNU Lesser General Public License along with this library; if not, write to the Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA Contact: todos@geneura.ugr.es, http://geneura.ugr.es Marc.Schoenauer@polytechnique.fr mkeijzer@dhi.dk */ //----------------------------------------------------------------------------- #ifndef _eoStat_h #define _eoStat_h #include // accumulate #include #include #include #include /** Base class for all statistics that need to be calculated over the (unsorted) population (I guess it is not really necessary? MS. Depstd::ends, there might be reasons to have a stat that is not an eoValueParam, but maybe I'm just kidding myself, MK) */ template class eoStatBase : public eoUF&, void> { public: virtual void lastCall(const eoPop&) {} }; /** The actual class that will be used as base for all statistics that need to be calculated over the (unsorted) population It is an eoStatBase AND an eoValueParam so it can be used in Monitors. */ template class eoStat : public eoValueParam, public eoStatBase { public : eoStat(T _value, std::string _description) : eoValueParam(_value, _description) {} }; /** Base class for statistics calculated over a sorted snapshot of the population */ template class eoSortedStatBase : public eoUF&, void> { public: virtual void lastCall(const std::vector&) {} }; /** The actual class that will be used as base for all statistics that need to be calculated over the sorted population It's an eoSortedStatBase AND an eoValueParam so it can be used in Monitors. */ template class eoSortedStat : public eoSortedStatBase, public eoValueParam { public : eoSortedStat(ParamType _value, std::string _desc) : eoValueParam(_value, _desc) {} }; /** Average fitness of a population. Fitness can be: - double - eoMinimizingFitness or eoMaximizingFitness - eoParetoFitness: The average of each objective is evaluated. ( For eoScalarFitnessAssembled user eoAssembledFitnessStat classes.) */ #ifdef _MSC_VER template class eoAverageStat : public eoStat #else template class eoAverageStat : public eoStat #endif { public : typedef typename EOT::Fitness Fitness; eoAverageStat(std::string _description = "Average Fitness") : eoStat(Fitness(), _description) {} static Fitness sumFitness(double _sum, const EOT& _eot){ _sum += _eot.fitness(); return _sum; } eoAverageStat(double _value, std::string _desc) : eoStat(_value, _desc) {} virtual void operator()(const eoPop& _pop){ doit(_pop, Fitness()); // specializations for scalar and std::vector } private : // Specialization for pareto fitness template void doit(const eoPop& _pop, eoParetoFitness) { value().clear(); value().resize(_pop[0].fitness().size(), 0.0); for (unsigned o = 0; o < value().size(); ++o) { for (unsigned i = 0; i < _pop.size(); ++i) { value()[o] += _pop[i].fitness()[o]; } value()[o] /= _pop.size(); } } // Default behavior template void doit(const eoPop& _pop, T) { Fitness v = std::accumulate(_pop.begin(), _pop.end(), Fitness(0.0), eoAverageStat::sumFitness); value() = v / _pop.size(); } }; /** Average fitness + Std. dev. of a population, fitness needs to be scalar. */ template class eoSecondMomentStats : public eoStat > { public : typedef typename EOT::Fitness fitness_type; typedef std::pair SquarePair; eoSecondMomentStats(std::string _description = "Average & Stdev") : eoStat(std::make_pair(0.0,0.0), _description) {} static SquarePair sumOfSquares(SquarePair _sq, const EOT& _eo) { double fitness = _eo.fitness(); _sq.first += fitness; _sq.second += fitness * fitness; return _sq; } virtual void operator()(const eoPop& _pop) { SquarePair result = std::accumulate(_pop.begin(), _pop.end(), std::make_pair(0.0, 0.0), eoSecondMomentStats::sumOfSquares); double n = _pop.size(); value().first = result.first / n; // average value().second = sqrt( (result.second - n * value().first * value().first) / (n - 1.0)); // stdev } }; /** The n_th element fitness in the population (see eoBestFitnessStat) */ #ifdef _MSC_VER template class eoNthElementFitnessStat : public eoSortedStat #else template class eoNthElementFitnessStat : public eoSortedStat #endif { public : typedef typename EOT::Fitness Fitness; eoNthElementFitnessStat(unsigned _whichElement, std::string _description = "nth element fitness") : eoSortedStat(Fitness(), _description), whichElement(_whichElement) {} virtual void operator()(const std::vector& _pop) { if (whichElement > _pop.size()) throw std::logic_error("fitness requested of element outside of pop"); doit(_pop, Fitness()); } private : struct CmpFitness { CmpFitness(unsigned _whichElement, bool _maxim) : whichElement(_whichElement), maxim(_maxim) {} bool operator()(const EOT* a, const EOT* b) { if (maxim) return a->fitness()[whichElement] > b->fitness()[whichElement]; return a->fitness()[whichElement] < b->fitness()[whichElement]; } unsigned whichElement; bool maxim; }; // Specialization for eoParetoFitness template void doit(const eoPop& _pop, eoParetoFitness) { typedef typename EOT::Fitness::fitness_traits traits; value().resize(traits::nObjectives()); // copy of pointers, what the heck std::vector tmp_pop = _pop; for (unsigned o = 0; o < value().size(); ++o) { typename std::vector::iterator nth = tmp_pop.begin() + whichElement; std::nth_element(tmp_pop.begin(), nth, tmp_pop.end(), CmpFitness(o, traits::maximizing(o))); value()[o] = (*nth)->fitness()[o]; } } // for everything else template void doit(const std::vector& _pop, T) { value() = _pop[whichElement]->fitness(); } unsigned whichElement; }; /* Actually, you shouldn't need to sort the population to get the best fitness MS - 17/11/00 But then again, if another stat needs sorted fitness anyway, getting the best out would be very fast. MK - 09/01/03 template class eoBestFitnessStat : public eoStat { public : typedef typename EOT::Fitness Fitness; eoBestFitnessStat(std::string _description = "Best Fitness") : eoStat(Fitness(), _description) {} virtual void operator()(const eoPop& _pop) { value() = _pop.nth_element_fitness(0); } }; */ /** Best fitness of a population. Fitness can be: - double - eoMinimizingFitness or eoMaximizingFitness - eoParetoFitness: ( For eoScalarFitnessAssembled look at eoAssembledFitnessStat ) */ #ifdef _MSC_VER template class eoBestFitnessStat : public eoStat #else template class eoBestFitnessStat : public eoStat #endif { public : typedef typename EOT::Fitness Fitness; eoBestFitnessStat(std::string _description = "Best ") : eoStat(Fitness(), _description) {} void operator()(const eoPop& _pop){ doit(_pop, Fitness() ); // specializations for scalar and std::vector } private : struct CmpFitness { CmpFitness(unsigned _which, bool _maxim) : which(_which), maxim(_maxim) {} bool operator()(const EOT& a, const EOT& b) { if (maxim) return a.fitness()[which] < b.fitness()[which]; return a.fitness()[which] > b.fitness()[which]; } unsigned which; bool maxim; }; // Specialization for pareto fitness template void doit(const eoPop& _pop, eoParetoFitness) { typedef typename EOT::Fitness::fitness_traits traits; value().resize(traits::nObjectives()); for (unsigned o = 0; o < traits::nObjectives(); ++o) { typename eoPop::const_iterator it = std::max_element(_pop.begin(), _pop.end(), CmpFitness(o, traits::maximizing(o))); value()[o] = it->fitness()[o]; } } // default template void doit(const eoPop& _pop, T) { // find the largest elements value() = _pop.best_element().fitness(); } }; template class eoDistanceStat : public eoStat { public : eoDistanceStat(std::string _name = "distance") : eoStat(0.0, _name) {} template double distance(T a, T b) { T res = a-b; return res < 0? -res : res; } double distance(bool a, bool b) { return (a==b)? 0 : 1; } void operator()(const eoPop& _pop) { double& v = value(); v = 0.0; for (unsigned i = 0; i < _pop.size(); ++i) { for (unsigned j = 0; j < _pop.size(); ++j) { for (unsigned k = 0; k < _pop[i].size(); ++k) { v += distance(_pop[i][k], _pop[j][k]); } } } double sz = _pop.size(); v /= sz * sz * _pop[0].size(); } }; /* template class eoStdevStat : public eoStat { public : typedef typename eoSecondMomentStats::SquarePair SquarePair; eoStdevStat(std::string _description = "Stdev") : eoStat(0.0, _description) {} virtual void operator()(const eoPop& _pop) { SquarePair result = std::accumulate(pop.begin(), pop.end(), std::make_pair(0.0, 0.0), eoSecondMomentStats::sumOfSquares); double n = pop.size(); value() = sqrt( (result.second - (result.first / n)) / (n - 1.0)); // stdev } }; */ #endif