nojhan/paradiseo
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

tutorial PEO modified

Browse source 
git-svn-id: svn://scm.gforge.inria.fr/svnroot/paradiseo@903 331e1502-861f-0410-8da2-ba01fb791d7f
This commit is contained in:
canape 2008-01-24 14:52:36 +00:00
commit 15db5e4fea
37 changed files with 512 additions and 832 deletions
Download patch file Download diff file Expand all files Collapse all files

75
trunk/paradiseo-peo/tutorial/Lesson3/mainEA.cpp
View file

@ -71,77 +71,72 @@ int main (int __argc, char *__argv[])
// First algorithm
/*****************************************************************************************/
eoGenContinue < Indi > genContPara (MAX_GEN);
eoCombinedContinue <Indi> continuatorPara (genContPara);
eoCheckPoint<Indi> checkpoint(continuatorPara);
peoEvalFunc<Indi> plainEval(f);
peoSeqPopEval< Indi > eval(plainEval); // Here, the evaluation is sequential
peoEvalFunc<Indi> mainEval( f );
peoPopEval <Indi> eval(mainEval);
eoUniformGenerator < double >uGen (INIT_POSITION_MIN, INIT_POSITION_MAX);
eoInitFixedLength < Indi > random (VEC_SIZE, uGen);
eoRankingSelect<Indi> selectionStrategy;
eoSelectNumber<Indi> select(selectionStrategy,POP_SIZE);
eoSegmentCrossover<Indi> crossover;
eoUniformMutation<Indi> mutation(EPSILON);
eoSGATransform<Indi> transform(crossover,CROSS_RATE,mutation,MUT_RATE);
peoSeqTransform<Indi> eaTransform(transform); // Here, the transformation is sequential
eoUniformMutation<Indi> mutation(EPSILON);
peoTransform<Indi> transform(crossover,CROSS_RATE,mutation,MUT_RATE);
peoPop < Indi > pop;
pop.append (POP_SIZE, random);
eoPlusReplacement<Indi> replace;
eoPop < Indi > pop;
pop.append (POP_SIZE, random);
eoPeriodicContinue <Indi> mig_cont( MIG_FREQ ); // Migration occurs periodically
eoRandomSelect<Indi> mig_select_one; // Emigrants are randomly selected
eoSelectNumber<Indi> mig_select (mig_select_one,MIG_SIZE);
eoPlusReplacement<Indi> mig_replace; // Immigrants replace the worse individuals
eoRandomSelect<Indi> mig_select_one;
eoSelector <Indi, peoPop<Indi> > mig_select (mig_select_one,MIG_SIZE,pop);
eoReplace <Indi, peoPop<Indi> > mig_replace (replace,pop);
peoSyncIslandMig<Indi, peoPop<Indi> > mig(MIG_FREQ,mig_select,mig_replace,topology,pop,pop);
checkpoint.add(mig);
eoEasyEA< Indi > eaAlg( checkpoint, eval, select, transform, replace );
peoWrapper parallelEA( eaAlg, pop);
eval.setOwner(parallelEA);
transform.setOwner(parallelEA);
mig.setOwner(parallelEA);
/*****************************************************************************************/
// Second algorithm (on the same model but with others names)
/*****************************************************************************************/
eoGenContinue < Indi > genContPara2 (MAX_GEN);
eoCombinedContinue <Indi> continuatorPara2 (genContPara2);
eoCheckPoint<Indi> checkpoint2(continuatorPara2);
peoEvalFunc<Indi> plainEval2(f);
peoSeqPopEval< Indi > eval2(plainEval2);
peoEvalFunc<Indi> mainEval2( f );
peoPopEval <Indi> eval2(mainEval2);
eoUniformGenerator < double >uGen2 (INIT_POSITION_MIN, INIT_POSITION_MAX);
eoInitFixedLength < Indi > random2 (VEC_SIZE, uGen2);
eoRankingSelect<Indi> selectionStrategy2;
eoSelectNumber<Indi> select2(selectionStrategy2,POP_SIZE);
eoSegmentCrossover<Indi> crossover2;
eoUniformMutation<Indi> mutation2(EPSILON);
eoSGATransform<Indi> transform2(crossover2,CROSS_RATE,mutation2,MUT_RATE);
peoSeqTransform<Indi> eaTransform2(transform2);
eoUniformMutation<Indi> mutation2(EPSILON);
peoTransform<Indi> transform2(crossover2,CROSS_RATE,mutation2,MUT_RATE);
peoPop < Indi > pop2;
pop2.append (POP_SIZE, random2);
eoPlusReplacement<Indi> replace2;
eoPop < Indi > pop2;
pop2.append (POP_SIZE, random2);
eoPeriodicContinue <Indi> mig_cont2( MIG_FREQ );
eoRandomSelect<Indi> mig_select_one2;
eoSelectNumber<Indi> mig_select2 (mig_select_one2,MIG_SIZE);
eoPlusReplacement<Indi> mig_replace2;
eoSelector <Indi, peoPop<Indi> > mig_select2 (mig_select_one2,MIG_SIZE,pop2);
eoReplace <Indi, peoPop<Indi> > mig_replace2 (replace2,pop2);
peoSyncIslandMig<Indi, peoPop<Indi> > mig2(MIG_FREQ,mig_select2,mig_replace2,topology,pop2,pop2);
checkpoint2.add(mig2);
eoEasyEA< Indi > eaAlg2( checkpoint2, eval2, select2, transform2, replace2 );
peoWrapper parallelEA2( eaAlg2, pop2);
eval2.setOwner(parallelEA2);
transform2.setOwner(parallelEA2);
mig2.setOwner(parallelEA2);
/*****************************************************************************************/
// You can choose between :
//
// - Synchronous communication : peoSyncIslandMig<Indi> mig(MIG_FREQ,mig_select,mig_replace,topology,pop,pop);
// - Asynchronous communication : peoAsyncIslandMig<Indi> mig(mig_cont,mig_select,mig_replace,topology,pop,pop);
// With a grid, you should use an asynchronous communication
peoAsyncIslandMig<Indi> mig(mig_cont,mig_select,mig_replace,topology,pop,pop2);
checkpoint.add(mig);
peoAsyncIslandMig<Indi> mig2(mig_cont2,mig_select2,mig_replace2,topology,pop2,pop);
checkpoint2.add(mig2);
// Initialization of the algorithms
peoEA<Indi> Algo(checkpoint,eval,select,eaTransform,replace);
mig.setOwner(Algo);
Algo(pop);
peoEA<Indi> Algo2(checkpoint2,eval2,select2,eaTransform2,replace2);
mig2.setOwner(Algo2);
Algo2(pop2);
peo :: run();
peo :: finalize();
if (getNodeRank()==1)
{
pop.sort();
pop2.sort();
std::cout << "Final population 1 :\n" << pop << std::endl;
std::cout << "Final population 2 :\n" << pop2 << std::endl;
}