Package org.moeaframework.core.operator.real

Class AdaptiveMetropolis

java.lang.Object

org.moeaframework.core.operator.real.MultiParentVariation

org.moeaframework.core.operator.real.AdaptiveMetropolis

All Implemented Interfaces:

Configurable, Variation

public class AdaptiveMetropolis
extends MultiParentVariation

The adaptive metropolis (AM) operator. AM is a multiparent operator, allowing a user-defined number of parents and offspring. AM produces normally-distributed clusters around each parent, where the shape of the distribution is controlled by the covariance of the parents.

Internally, the Cholesky decomposition is used to update the resulting offspring distribution. Cholesky decomposition requires that its input be positive definite. In order to guarantee this condition is satisfied, all parents must be unique. In the event that the positive definite condition is not satisfied, no offspring are produced and an empty array is returned by evolve(Solution[]).

References:

1. Vrugt, J.A., Robinson, B.A. and Hyman, J.M. "Self-Adaptive MultiMethod Search For Global Optimization in Real-Parameter Spaces." IEEE Transactions on Evolutionary Computation, pp. 1-17, 2009.
2. Vrugt, J.A. and Robinson, B.A. "Improved Evolutionary Optimization from Genetically Adaptive Multimethod Search." Proceedings of the National Academy of Sciences of the United States of America, vol. 104, pp. 708 - 711, 2007.
3. Gelman, A., Roberts, G.O. and Gilks, W.R. "Efficient Metropolis Jumping Rules." Bayesian Statistics, vol. 5, pp. 599-607, 1996.

Field Summary

Fields inherited from class org.moeaframework.core.operator.real.MultiParentVariation

numberOfOffspring, numberOfParents

Constructor Summary

Constructors

Constructor	Description
AdaptiveMetropolis()	Constructs an adaptive metropolis operator with default settings, taking 10 parents and producing 2 offspring.
AdaptiveMetropolis(int numberOfParents, int numberOfOffspring, double jumpRateCoefficient)	Constructs an adaptive metropolis operator.

Method Summary

Modifier and Type	Method	Description
Solution[]	evolve(Solution[] parents)	Evolves one or more parent solutions (specified by getArity) and produces one or more child solutions.
double	getJumpRateCoefficient()	Returns the jump rate coefficient value.
String	getName()	Returns the name of this variation operator.
void	setJumpRateCoefficient(double jumpRateCoefficient)	Sets the jump rate coefficient, controlling the standard deviation of the covariance matrix.

Methods inherited from class org.moeaframework.core.operator.real.MultiParentVariation

getArity, getNumberOfOffspring, getNumberOfParents, setNumberOfOffspring, setNumberOfParents

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.configuration.Configurable

applyConfiguration, getConfiguration

Constructor Details

AdaptiveMetropolis

public AdaptiveMetropolis()

Constructs an adaptive metropolis operator with default settings, taking 10 parents and producing 2 offspring. The recommended value of 2.4 is used for the jump rate coefficient.

AdaptiveMetropolis

public AdaptiveMetropolis(int numberOfParents,
 int numberOfOffspring,
 double jumpRateCoefficient)

Constructs an adaptive metropolis operator.

Parameters:

numberOfParents - the number of parents required by this operator

numberOfOffspring - the number of parents produced by this operator

jumpRateCoefficient - the jump raote coefficient, controlling the standard deviation of the covariance matrix

Method Details

getName

public String getName()

Description copied from interface: Variation

Returns the name of this variation operator. This name should also be used as the prefix for any parameters. As such, the name should only contain alphanumeric characters, avoid using whitespace and other symbols.

Returns:

the name of this variation operator

getJumpRateCoefficient

public double getJumpRateCoefficient()

Returns the jump rate coefficient value.

Returns:

the jump rate coefficient value

setJumpRateCoefficient

public void setJumpRateCoefficient(double jumpRateCoefficient)

Sets the jump rate coefficient, controlling the standard deviation of the covariance matrix. The actual jump rate is calculated as Math.pow(jumpRateCoefficient / Math.sqrt(n), 2.0), where n is the number of decision variables. The recommended value is 2.4.

Parameters:

jumpRateCoefficient - the jump rate coefficient value

evolve

public Solution[] evolve(Solution[] parents)

Description copied from interface: Variation

Evolves one or more parent solutions (specified by getArity) and produces one or more child solutions. By contract, the parents must not be modified. The copy constructor should be used to create copies of the parents with these copies subsequently modified.

Parameters:

parents - the array of parent solutions

Returns:

an array of child solutions