org.moeaframework.algorithm
Class AdaptiveTimeContinuation

java.lang.Object
  org.moeaframework.algorithm.PeriodicAction
      org.moeaframework.algorithm.AdaptiveTimeContinuation

All Implemented Interfaces:

Algorithm, EvolutionaryAlgorithm

Direct Known Subclasses:

EpsilonProgressContinuation

public class AdaptiveTimeContinuation
extends PeriodicAction
implements EvolutionaryAlgorithm

Decorator for EvolutionaryAlgorithms to add time continuation (restarts). Restarts occur if either

1. the number of fitness function evaluations since the last restart exceeds maxWindowSize; or
2. the population-to-archive ratio exceeds populationRatio by more than 25%.

If a restart occurs, the population is emptied, the population size is adapted to maintain the populationRatio, the the new population is filled with solutions selected from algorithm.getArchive() and mutated using the specified Selection and Variation operators.

References:

1. Goldberg, D. E. "Sizing Populations for Serial and Parallel Genetic Algorithms." In 3rd International Conference on Genetic Algorithms, pp. 70-79, 1989.
2. Srivastava, R. P. "Time Continuation in Genetic Algorithms." Technical report, Illinois Genetic Algorithm Laboratory, 2002.
3. Kollat, J. B., and Reed, P. M. "Comparison of Multi-Objective Evolutionary Algorithms for Long-Term Monitoring Design." Advances in Water Resources, 29(6):792-807, 2006.

Nested Class Summary

Nested classes/interfaces inherited from class org.moeaframework.algorithm.PeriodicAction

PeriodicAction.FrequencyType

Field Summary

Fields inherited from class org.moeaframework.algorithm.PeriodicAction

algorithm, frequency, frequencyType, iteration, lastInvocation

Constructor Summary

AdaptiveTimeContinuation(EvolutionaryAlgorithm algorithm, int windowSize, int maxWindowSize, double populationRatio, int minimumPopulationSize, int maximumPopulationSize, Selection selection, Variation variation)
Decorates the specified algorithm with adaptive time continuation.

Method Summary

void	addRestartListener(RestartListener listener) Adds a listener to be notified whenever a restart occurs.
protected RestartType	check() Performs a check to determine if a restart should occur.
void	doAction() Invoked periodically by this class to perform some function.
NondominatedPopulation	getArchive() Returns the current non-dominated archive of the best solutions generated by this evolutionary algorithm, or null if no archive is used.
Population	getPopulation() Returns the current population of this evolutionary algorithm.
Serializable	getState() Returns a Serializable object representing the internal state of this algorithm.
void	removeRestartListener(RestartListener listener) Removes the specified listener so it no longer receives notifications whenever a restart occurs.
protected void	restart(RestartType type) Performs a restart.
void	setState(Object objState) Sets the internal state of of this algorithm.

Methods inherited from class org.moeaframework.algorithm.PeriodicAction

evaluate, getNumberOfEvaluations, getProblem, getResult, isTerminated, step, terminate

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.moeaframework.core.Algorithm

evaluate, getNumberOfEvaluations, getProblem, getResult, isTerminated, step, terminate

Constructor Detail

AdaptiveTimeContinuation

public AdaptiveTimeContinuation(EvolutionaryAlgorithm algorithm,
                                int windowSize,
                                int maxWindowSize,
                                double populationRatio,
                                int minimumPopulationSize,
                                int maximumPopulationSize,
                                Selection selection,
                                Variation variation)

Decorates the specified algorithm with adaptive time continuation.

Parameters:

algorithm - the algorithm being decorated

windowSize - the number of iterations between invocations of check

maxWindowSize - the maximum number of iterations allowed since the last restart before forcing a restart

populationRatio - the population-to-archive ratio

minimumPopulationSize - the minimum size of the population

maximumPopulationSize - the maximum size of the population

selection - the selection operator for selecting solutions from the archive during a restart

variation - the variation operator for mutating solutions selected from the archive during a restart

Method Detail

addRestartListener

public void addRestartListener(RestartListener listener)

Adds a listener to be notified whenever a restart occurs.

Parameters:

listener - the listener to be notified whenever a restart occurs

removeRestartListener

public void removeRestartListener(RestartListener listener)

Removes the specified listener so it no longer receives notifications whenever a restart occurs.

Parameters:

listener - the listener to be removed

check

protected RestartType check()

Performs a check to determine if a restart should occur. Returns RestartType.NONE if no restart should occur; or RestartType.HARD if the population-to-archive ratio exceeds populationRatio by more than 25% or the number of fitness evaluations since the last restart exceeds maxWindowSize.

Returns:

RestartType.NONE if no restart should occur; or RestartType.HARD if the population-to-archive ratio exceeds populationRatio by more than 25% or if the number of fitness evaluations since the last restart exceeds maxWindowSize

restart

protected void restart(RestartType type)

Performs a restart. If the type is RestartType.HARD, the population is emptied, resized and filled with solutions selected and mutated from the archive. If the type is RestartType.SOFT, the population is not emptied; new solutions are only added to fill any empty slots.

Parameters:

type - the type of restart

doAction

public void doAction()

Description copied from class: PeriodicAction

Invoked periodically by this class to perform some function. This function should not depend on the frequency that it is invoked, since it is not guaranteed that this method is invoked at any specific frequency and, in some cases, may not be invoked at all.

Specified by:

doAction in class PeriodicAction

getPopulation

public Population getPopulation()

Description copied from interface: EvolutionaryAlgorithm

Returns the current population of this evolutionary algorithm.

Specified by:

getPopulation in interface EvolutionaryAlgorithm

Returns:

the current population of this evolutionary algorithm

getArchive

public NondominatedPopulation getArchive()

Description copied from interface: EvolutionaryAlgorithm

Returns the current non-dominated archive of the best solutions generated by this evolutionary algorithm, or null if no archive is used.

Specified by:

getArchive in interface EvolutionaryAlgorithm

Returns:

the current non-dominated archive of the best solutions generated by this evolutionary algorithm, or null if no archive is used

getState

public Serializable getState()
                      throws NotSerializableException

Description copied from interface: Algorithm

Returns a Serializable object representing the internal state of this algorithm.

Specified by:

getState in interface Algorithm

Overrides:

getState in class PeriodicAction

Returns:

a Serializable object representing the internal state of this algorithm

Throws:

NotSerializableException - if this algorithm does not support serialization

setState

public void setState(Object objState)
              throws NotSerializableException

Description copied from interface: Algorithm

Sets the internal state of of this algorithm.

Specified by:

setState in interface Algorithm

Overrides:

setState in class PeriodicAction

Parameters:

objState - the internal state of this algorithm

Throws:

NotSerializableException - if this algorithm does not support serialization