Ajout du moHillClimberSampling, et ajout des méthodes init dans les stats ;)

git-svn-id: svn://scm.gforge.inria.fr/svnroot/paradiseo@1787 331e1502-861f-0410-8da2-ba01fb791d7f
This commit is contained in:
verel 2010-05-05 12:42:39 +00:00
commit c46fac7da4
9 changed files with 330 additions and 12 deletions

View file

@ -58,7 +58,7 @@ public:
*/
moRandomSearch(eoInit<EOT> & _init, eoEvalFunc<EOT>& _fullEval, unsigned _nbSolMax):
moLocalSearch<Neighbor>(explorer, trueCont, _fullEval),
explorer(_init, _fullEval, _nbSolMax)
explorer(_init, _fullEval, _nbSolMax - 1)
{}
/**
@ -69,7 +69,7 @@ public:
*/
moRandomSearch(eoInit<EOT> & _init, eoEvalFunc<EOT>& _fullEval, unsigned _nbSolMax, moContinuator<Neighbor>& _cont):
moLocalSearch<Neighbor>(explorer, _cont, _fullEval),
explorer(_init, _fullEval, _nbSolMax)
explorer(_init, _fullEval, _nbSolMax - 1)
{}
private:

View file

@ -137,6 +137,7 @@
#include <sampling/moSampling.h>
#include <sampling/moDensityOfStatesSampling.h>
#include <sampling/moAutocorrelationSampling.h>
#include <sampling/moHillClimberSampling.h>
#include <problems/bitString/moBitNeighbor.h>
#include <problems/eval/moOneMaxIncrEval.h>

View file

@ -54,8 +54,8 @@ public:
* @param _solution to perturb
*/
void operator()(EOT& _solution){
init(solution);
ls(solution);
init(_solution);
ls(_solution);
}
private:

View file

@ -43,10 +43,9 @@
#include <sampling/moSampling.h>
/**
* To compute the autocorrelation function:
* Perform a random walk based on the neighborhood,
* The fitness values of solutions are collected during the random walk
* The autocorrelation can be computed from the serie of fitness values
* To compute the density of states:
* Sample the fitness of random solution in the search space
* The fitness values of solutions are collected during the random search
*
*/
template <class Neighbor>

View file

@ -0,0 +1,107 @@
/*
<moHillClimberSampling.h>
Copyright (C) DOLPHIN Project-Team, INRIA Lille - Nord Europe, 2006-2010
Sebastien Verel, Arnaud Liefooghe, Jeremie Humeau
This software is governed by the CeCILL license under French law and
abiding by the rules of distribution of free software. You can use,
modify and/ or redistribute the software under the terms of the CeCILL
license as circulated by CEA, CNRS and INRIA at the following URL
"http://www.cecill.info".
As a counterpart to the access to the source code and rights to copy,
modify and redistribute granted by the license, users are provided only
with a limited warranty and the software's author, the holder of the
economic rights, and the successive licensors have only limited liability.
In this respect, the user's attention is drawn to the risks associated
with loading, using, modifying and/or developing or reproducing the
software by the user in light of its specific status of free software,
that may mean that it is complicated to manipulate, and that also
therefore means that it is reserved for developers and experienced
professionals having in-depth computer knowledge. Users are therefore
encouraged to load and test the software's suitability as regards their
requirements in conditions enabling the security of their systems and/or
data to be ensured and, more generally, to use and operate it in the
same conditions as regards security.
The fact that you are presently reading this means that you have had
knowledge of the CeCILL license and that you accept its terms.
ParadisEO WebSite : http://paradiseo.gforge.inria.fr
Contact: paradiseo-help@lists.gforge.inria.fr
*/
#ifndef moHillClimberSampling_h
#define moHillClimberSampling_h
#include <eoInit.h>
#include <eval/moEval.h>
#include <eoEvalFunc.h>
#include <continuator/moCheckpoint.h>
#include <perturb/moLocalSearchInit.h>
#include <algo/moRandomSearch.h>
#include <algo/moSimpleHC.h>
#include <continuator/moSolutionStat.h>
#include <continuator/moCounterStat.h>
#include <continuator/moStatFromStat.h>
#include <sampling/moSampling.h>
/**
* To compute the length and final solution of an adaptive walk:
* Perform a simple Hill-climber based on the neighborhood (adaptive walk),
* The lengths of HC are collected and the final solution which are local optima
* The adaptive walk is repeated several times
*
*/
template <class Neighbor>
class moHillClimberSampling : public moSampling<Neighbor>
{
public:
typedef typename Neighbor::EOT EOT ;
using moSampling<Neighbor>::localSearch;
/**
* Default Constructor
* @param _init initialisation method of the solution
* @param _neighborhood neighborhood giving neighbor in random order
* @param _nbAdaptWalk Number of adaptive walks
*/
moHillClimberSampling(eoInit<EOT> & _init,
moNeighborhood<Neighbor> & _neighborhood,
eoEvalFunc<EOT>& _fullEval, moEval<Neighbor>& _eval,
unsigned int _nbAdaptWalk) :
moSampling<Neighbor>(initHC, * new moRandomSearch<Neighbor>(initHC, _fullEval, _nbAdaptWalk), copyStat),
copyStat(lengthStat),
checkpoint(trueCont),
hc(_neighborhood, _fullEval, _eval, checkpoint),
initHC(_init, hc)
{
// to count the number of step in the HC
checkpoint.add(lengthStat);
// add the solution into statistics
add(solStat);
}
/**
* default destructor
*/
~moHillClimberSampling() {
// delete the pointer on the local search which has been constructed in the constructor
delete &localSearch;
}
protected:
moSolutionStat<EOT> solStat;
moCounterStat<EOT> lengthStat;
moTrueContinuator<Neighbor> trueCont;
moStatFromStat<EOT, unsigned int> copyStat;
moCheckpoint<Neighbor> checkpoint;
moSimpleHC<Neighbor> hc;
moLocalSearchInit<Neighbor> initHC;
};
#endif

View file

@ -102,7 +102,7 @@ public:
* the statistics are stored in the moVectorMonitor vector
*/
void operator()(void) {
// clear all statisic vectors
// clear all statistic vectors
for(unsigned i = 0; i < monitorVec.size(); i++)
monitorVec[i]->clear();

View file

@ -12,6 +12,7 @@ ADD_EXECUTABLE(testRandomNeutralWalk testRandomNeutralWalk.cpp)
ADD_EXECUTABLE(sampling sampling.cpp)
ADD_EXECUTABLE(densityOfStates densityOfStates.cpp)
ADD_EXECUTABLE(autocorrelation autocorrelation.cpp)
ADD_EXECUTABLE(adaptiveWalks adaptiveWalks.cpp)
TARGET_LINK_LIBRARIES(testRandomWalk eoutils ga eo)
TARGET_LINK_LIBRARIES(testMetropolisHasting eoutils ga eo)
@ -19,3 +20,4 @@ TARGET_LINK_LIBRARIES(testRandomNeutralWalk eoutils ga eo)
TARGET_LINK_LIBRARIES(sampling eoutils ga eo)
TARGET_LINK_LIBRARIES(densityOfStates eoutils ga eo)
TARGET_LINK_LIBRARIES(autocorrelation eoutils ga eo)
TARGET_LINK_LIBRARIES(adaptiveWalks eoutils ga eo)

View file

@ -0,0 +1,209 @@
//-----------------------------------------------------------------------------
/** adaptiveWalks.cpp
*
* SV - 05/05/10
*
*/
//-----------------------------------------------------------------------------
// standard includes
#define HAVE_SSTREAM
#include <stdexcept> // runtime_error
#include <iostream> // cout
#include <sstream> // ostrstream, istrstream
#include <fstream>
#include <string.h>
// the general include for eo
#include <eo>
// declaration of the namespace
using namespace std;
//-----------------------------------------------------------------------------
// representation of solutions, and neighbors
#include <ga/eoBit.h> // bit string : see also EO tutorial lesson 1: FirstBitGA.cpp
#include <problems/bitString/moBitNeighbor.h> // neighbor of bit string
//-----------------------------------------------------------------------------
// fitness function, and evaluation of neighbors
#include <eval/oneMaxEval.h>
#include <problems/eval/moOneMaxIncrEval.h>
#include <eval/moFullEvalByModif.h>
//-----------------------------------------------------------------------------
// neighborhood description
#include <neighborhood/moOrderNeighborhood.h> // visit all the neighbors
//-----------------------------------------------------------------------------
// the sampling class
#include <sampling/moHillClimberSampling.h>
// Declaration of types
//-----------------------------------------------------------------------------
// Indi is the typedef of the solution type like in paradisEO-eo
typedef eoBit<unsigned int> Indi; // bit string with unsigned fitness type
// Neighbor is the typedef of the neighbor type,
// Neighbor = How to compute the neighbor from the solution + information on it (i.e. fitness)
// all classes from paradisEO-mo use this template type
typedef moBitNeighbor<unsigned int> Neighbor ; // bit string neighbor with unsigned fitness type
void main_function(int argc, char **argv)
{
/* =========================================================
*
* Parameters
*
* ========================================================= */
// more information on the input parameters: see EO tutorial lesson 3
// but don't care at first it just read the parameters of the bit string size and the random seed.
// First define a parser from the command-line arguments
eoParser parser(argc, argv);
// For each parameter, define Parameter, read it through the parser,
// and assign the value to the variable
// random seed parameter
eoValueParam<uint32_t> seedParam(time(0), "seed", "Random number seed", 'S');
parser.processParam( seedParam );
unsigned seed = seedParam.value();
// length of the bit string
eoValueParam<unsigned int> vecSizeParam(20, "vecSize", "Genotype size", 'V');
parser.processParam( vecSizeParam, "Representation" );
unsigned vecSize = vecSizeParam.value();
// the number of adaptive walks
eoValueParam<unsigned int> solParam(100, "nbSol", "Number of adaptive walks", 'n');
parser.processParam( solParam, "Representation" );
unsigned nbSol = solParam.value();
// the name of the output file
string str_out = "out.dat"; // default value
eoValueParam<string> outParam(str_out.c_str(), "out", "Output file of the sampling", 'o');
parser.processParam(outParam, "Persistence" );
// the name of the "status" file where all actual parameter values will be saved
string str_status = parser.ProgramName() + ".status"; // default value
eoValueParam<string> statusParam(str_status.c_str(), "status", "Status file");
parser.processParam( statusParam, "Persistence" );
// do the following AFTER ALL PARAMETERS HAVE BEEN PROCESSED
// i.e. in case you need parameters somewhere else, postpone these
if (parser.userNeedsHelp()) {
parser.printHelp(cout);
exit(1);
}
if (statusParam.value() != "") {
ofstream os(statusParam.value().c_str());
os << parser;// and you can use that file as parameter file
}
/* =========================================================
*
* Random seed
*
* ========================================================= */
// reproducible random seed: if you don't change SEED above,
// you'll aways get the same result, NOT a random run
// more information: see EO tutorial lesson 1 (FirstBitGA.cpp)
rng.reseed(seed);
/* =========================================================
*
* Initialization of the solution
*
* ========================================================= */
// a Indi random initializer: each bit is random
// more information: see EO tutorial lesson 1 (FirstBitGA.cpp)
eoUniformGenerator<bool> uGen;
eoInitFixedLength<Indi> random(vecSize, uGen);
/* =========================================================
*
* Eval fitness function (full evaluation)
*
* ========================================================= */
// the fitness function is just the number of 1 in the bit string
oneMaxEval<Indi> fullEval;
/* =========================================================
*
* evaluation of a neighbor solution
*
* ========================================================= */
// Use it if there is no incremental evaluation: a neighbor is evaluated by the full evaluation of a solution
// moFullEvalByModif<Neighbor> neighborEval(fullEval);
// Incremental evaluation of the neighbor: fitness is modified by +/- 1
moOneMaxIncrEval<Neighbor> neighborEval;
/* =========================================================
*
* the neighborhood of a solution
*
* ========================================================= */
// Exploration of the neighborhood in order
// from bit 0 to bit vecSize-1
moOrderNeighborhood<Neighbor> neighborhood(vecSize);
/* =========================================================
*
* The sampling of the search space
*
* ========================================================= */
// sampling object :
// - random initialization
// - local search to sample the search space
// - one statistic to compute
moHillClimberSampling<Neighbor> sampling(random, neighborhood, fullEval, neighborEval, nbSol);
/* =========================================================
*
* execute the sampling
*
* ========================================================= */
sampling();
/* =========================================================
*
* export the sampling
*
* ========================================================= */
// to export the statistics into file
sampling.fileExport(str_out);
// to get the values of statistics
// so, you can compute some statistics in c++ from the data
const std::vector<double> & lengthValues = sampling.getValues(0);
std::cout << "First values:" << std::endl;
std::cout << "Length " << lengthValues[0] << std::endl;
std::cout << "Last values:" << std::endl;
std::cout << "Length " << lengthValues[lengthValues.size() - 1] << std::endl;
}
// A main that catches the exceptions
int main(int argc, char **argv)
{
try {
main_function(argc, argv);
}
catch (exception& e) {
cout << "Exception: " << e.what() << '\n';
}
return 1;
}

View file

@ -70,7 +70,7 @@ void main_function(int argc, char **argv)
parser.processParam( vecSizeParam, "Representation" );
unsigned vecSize = vecSizeParam.value();
// the number of steps of the random walk
// the number of solution sampled
eoValueParam<unsigned int> solParam(100, "nbSol", "Number of random solution", 'n');
parser.processParam( solParam, "Representation" );
unsigned nbSol = solParam.value();
@ -135,8 +135,8 @@ void main_function(int argc, char **argv)
// sampling object :
// - random initialization
// - local search to sample the search space
// - one statistic to compute
// - fitness function
// - number of solutions to sample
moDensityOfStatesSampling<Neighbor> sampling(random, fullEval, nbSol);
/* =========================================================