Package org.opt4j.optimizer.mopso

Class MOPSO

java.lang.Object

org.opt4j.core.optimizer.AbstractOptimizer

org.opt4j.optimizer.mopso.MOPSO

All Implemented Interfaces:

Optimizer

public class MOPSO
extends AbstractOptimizer

The MOPSO is an implementation of a multi-objective particle swarm optimizer. This implementation is based on the OMOPSO. Thus, this MOPSO is restricted to problems that are based on the DoubleGenotype.

See Also:

"Improving PSO-based Multi-Objective Optimization using Crowding, Mutation and e-Dominance, M. Reyes Sierra and C. A. Coello Coello, In Proceedings of Evolutionary Multi-Criterion Optimization, 2005"

Field Summary

Fields

Modifier and Type	Field	Description
protected AlgebraDouble	algebra
protected Archive	leaders
protected MutationRate	mutationRate
protected MutateDoubleUniform	nonUniform
protected ParticleFactory	particleFactory
protected Term	positionTerm
protected Random	random
protected int	size
protected MutateDoubleUniform	uniform
protected VelocityTerm	velocityTerm

Fields inherited from class org.opt4j.core.optimizer.AbstractOptimizer

archive, completer, control, iteration, iterationListeners, optimizing, population, stateListeners

Constructor Summary

Constructors

Constructor	Description
MOPSO(Population population, Archive archive, IndividualFactory individualFactory, IndividualCompleter completer, Control control, Rand random, MutateDoubleUniform uniform, MutateDoubleNonUniform nonUniform, MutationRate mutationRate, Iteration iteration, int size, int archiveSize)	Constructs a MOPSO.

Method Summary

Modifier and Type	Method	Description
protected boolean	dominates(Particle old, Particle current)	Returns true if the old particle position dominates the new one.
protected Map<Particle,Particle>	getLeaders(Archive leaders, Population population)	Assigns each Particle a leader.
protected void	initPopulation()	Initialize the Population.
protected Map<Particle,Particle>	move(Population population, Map<Particle,Particle> leaders)	Determine the new positions for the Particles in the Population.
void	optimize()	Starts the optimization process.
protected void	updateLeaders(Archive leaders, Population population)	Update the global leaders Archive.
protected void	updatePersonalBest(Map<Particle,Particle> next)	Update the personal best of each Particle.

Methods inherited from class org.opt4j.core.optimizer.AbstractOptimizer

addOptimizerIterationListener, addOptimizerStateListener, getIteration, injectListeners, isRunning, nextIteration, removeOptimizerIterationListener, removeOptimizerStateListener, startOptimization, stopOptimization

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

size

protected final int size

leaders

protected final Archive leaders

particleFactory

protected final ParticleFactory particleFactory

algebra

protected final AlgebraDouble algebra

random

protected final Random random

velocityTerm

protected final VelocityTerm velocityTerm

positionTerm

protected final Term positionTerm

uniform

protected final MutateDoubleUniform uniform

nonUniform

protected final MutateDoubleUniform nonUniform

mutationRate

protected final MutationRate mutationRate

Constructor Detail

MOPSO

@Inject
public MOPSO(Population population,
             Archive archive,
             IndividualFactory individualFactory,
             IndividualCompleter completer,
             Control control,
             Rand random,
             MutateDoubleUniform uniform,
             MutateDoubleNonUniform nonUniform,
             MutationRate mutationRate,
             Iteration iteration,
             int size,
             int archiveSize)

Constructs a MOPSO.

Parameters:

population - the population

archive - the archive

individualFactory - the individual (particle) factory

completer - the completer

control - the control

random - the random number generator

uniform - the uniform mutation

nonUniform - the non-uniform mutation

mutationRate - the mutation rate

iteration - the iteration counter

size - the number of particles

archiveSize - the size of the archive for the global leaders

Method Detail

optimize

public void optimize()
              throws StopException,
                     TerminationException

Description copied from interface: Optimizer

Starts the optimization process.

Throws:

StopException - if the optimization is stopped

TerminationException - if the optimization is terminated

move

protected Map<Particle,Particle> move(Population population,
                                            Map<Particle,Particle> leaders)

Determine the new positions for the Particles in the Population.

Parameters:

population - the population

leaders - the map for the global leaders

Returns:

the map of the old to the new particle

initPopulation

protected void initPopulation()

Initialize the Population.

updateLeaders

protected void updateLeaders(Archive leaders,
                             Population population)

Update the global leaders Archive.

Parameters:

leaders - the archive

population - the population

updatePersonalBest

protected void updatePersonalBest(Map<Particle,Particle> next)

Update the personal best of each Particle.

Parameters:

next - the old and new positions

dominates

protected boolean dominates(Particle old,
                            Particle current)

Returns true if the old particle position dominates the new one.

Parameters:

old - the old particle

current - the new particle

Returns:

true if the old particle position dominates the new one

getLeaders

protected Map<Particle,Particle> getLeaders(Archive leaders,
                                                  Population population)

Assigns each Particle a leader.

Parameters:

leaders - the archive of available leader

population - the population

Returns:

the map of each particle to its leader