paradiseo/eo/src/eoRankingCached.h
2025-04-15 17:48:00 +02:00

139 lines
5 KiB
C++

/** -*- mode: c++; c-indent-level: 4; c++-member-init-indent: 8; comment-column: 35; -*-
-----------------------------------------------------------------------------
eoRankingCached.h
(c) Maarten Keijzer, Marc Schoenauer, 2001
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 eoRankingCached_h
#define eoRankingCached_h
#include "eoPerf2Worth.h"
/**
* @class eoRankingCached
* @brief Cached version of eoRanking that stores precomputed values for better performance
*
* This class implements the same ranking algorithm as eoRanking but adds a caching layer
* that stores frequently used values when the population size remains constant between
* calls. This optimization is particularly useful in steady-state evolution where the
* population size typically doesn't change between selection operations.
*
* The caching mechanism stores:
* - Population size related values (pSize, pSizeMinusOne)
* - Precomputed coefficients (alpha, beta, gamma)
*
* @warning This optimization should only be used when the population size remains constant
* between calls to the operator. For dynamic population sizes, use the standard eoRanking.
*
* @ingroup Selectors
*/
template <class EOT>
class eoRankingCached : public eoPerf2Worth<EOT>
{
public:
using eoPerf2Worth<EOT>::value;
/* Ctor:
@param _p selective pressure (in (1,2]
@param _e exponent (1 == linear)
*/
eoRankingCached(double _p = 2.0, double _e = 1.0) : pressure(_p), exponent(_e), cached_pSize(0)
{
assert(1 < pressure and pressure <= 2);
}
/*
Computes the ranked fitness with caching optimization
Fitnesses range in [m,M] where:
- m = 2-pressure/popSize
- M = pressure/popSize
The progression between m and M depends on the exponent (linear when exponent=1)
@param _pop The population to rank
*/
virtual void operator()(const eoPop<EOT> &_pop)
{
unsigned pSize = _pop.size();
if (pSize <= 1)
throw eoPopSizeException(pSize, "cannot do ranking with population of size <= 1");
// value() refers to the std::vector of worthes (we're in an eoParamvalue)
value().resize(pSize);
// Cache population-size dependent values only when population size changes
if (pSize != cached_pSize)
{
cached_pSize = pSize;
cached_pSizeMinusOne = pSize - 1;
cached_beta = (2 - pressure) / pSize;
cached_gamma = (2 * pressure - 2) / pSize;
cached_alpha = (2 * pressure - 2) / (pSize * cached_pSizeMinusOne);
}
std::vector<const EOT *> rank;
_pop.sort(rank);
// map of indices for the population
std::unordered_map<const EOT *, unsigned> indexMap;
for (unsigned i = 0; i < pSize; ++i)
{
indexMap[&_pop[i]] = i;
}
if (exponent == 1.0) // no need for exponential then (linear case)
{
for (unsigned i = 0; i < pSize; i++)
{
const EOT *indiv = rank[i];
int which = indexMap[indiv];
value()[which] = cached_alpha * (pSize - i) + cached_beta;
}
}
else // non-linear case (exponent != 1)
{
for (unsigned i = 0; i < pSize; i++)
{
const EOT *indiv = rank[i];
int which = indexMap[indiv];
// value is in [0,1]
double tmp = ((double)(pSize - i)) / pSize;
// to the exponent, and back to [m,M]
value()[which] = cached_gamma * pow(tmp, exponent) + cached_beta;
}
}
}
private:
double pressure; // selective pressure (1 < pressure <= 2)
double exponent; // exponent (1 = linear)
// Cached values (recomputed only when population size changes)
unsigned cached_pSize; // last seen population size
unsigned cached_pSizeMinusOne; // pSize - 1
double cached_alpha; // linear scaling coefficient
double cached_beta; // base value coefficient
double cached_gamma; // non-linear scaling coefficient
};
#endif