Package org.moeaframework.algorithm.single

Class VectorAngleDistanceScalingComparator

java.lang.Object

org.moeaframework.algorithm.single.VectorAngleDistanceScalingComparator

All Implemented Interfaces:

Comparator<Solution>, AggregateObjectiveComparator, DominanceComparator

public class VectorAngleDistanceScalingComparator
extends Object
implements AggregateObjectiveComparator

The vector angle distance scaling aggregate function. The distance between the origin and the objective vector is scaled by the cosine angle between the objective vector and a target vector.

References:

1. E. J. Hughes. "Multiple Single Objective Pareto Sampling." 2003 Congress on Evolutionary Computation, pp. 2678-2684.

Constructor Summary

Constructors

Constructor	Description
VectorAngleDistanceScalingComparator(double[] weights)	Constructs a new instance of the vector angle distance scaling aggregate function with the given weights.
VectorAngleDistanceScalingComparator(double[] weights, double q)	Constructs a new instance of the vector angle distance scaling aggregate function with the given weights.

Method Summary

Modifier and Type	Method	Description
double	calculateFitness(Solution solution)	Calculates the aggregate (fitness) value of the solution using this aggregate function.
static double	calculateFitness(Solution solution, double[] weights, double q)	Computes the vector angle distance scaling aggregate fitness of the solution.
int	compare(Solution solution1, Solution solution2)	Compares the two solutions using a dominance relation, returning -1 if solution1 dominates solution2, 1 if solution2 dominates solution1, and 0 if the solutions are non-dominated.
double	getAngleScalingFactor()	Returns the factor for scaling the effects of the angle.
double[]	getWeights()	Returns the weights used by this aggregate function.

Methods inherited from class java.lang.Object

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

Methods inherited from interface java.util.Comparator

equals, reversed, thenComparing, thenComparing, thenComparing, thenComparingDouble, thenComparingInt, thenComparingLong

Constructor Details

VectorAngleDistanceScalingComparator

public VectorAngleDistanceScalingComparator(double[] weights)

Constructs a new instance of the vector angle distance scaling aggregate function with the given weights.

Parameters:

weights - the weight vector; must have a magnitude of 1.0

VectorAngleDistanceScalingComparator

public VectorAngleDistanceScalingComparator(double[] weights,
 double q)

Constructs a new instance of the vector angle distance scaling aggregate function with the given weights.

Parameters:

weights - the weight vector; must have a magnitude of 1.0

q - factor for scaling the effects of the angle

Method Details

getAngleScalingFactor

public double getAngleScalingFactor()

Returns the factor for scaling the effects of the angle.

Returns:

the factor for scaling the effects of the angle

getWeights

public double[] getWeights()

Description copied from interface: AggregateObjectiveComparator

Returns the weights used by this aggregate function.

Specified by:

getWeights in interface AggregateObjectiveComparator

Returns:

the weights

compare

public int compare(Solution solution1,
 Solution solution2)

Description copied from interface: DominanceComparator

Compares the two solutions using a dominance relation, returning -1 if solution1 dominates solution2, 1 if solution2 dominates solution1, and 0 if the solutions are non-dominated.

Specified by:

compare in interface Comparator<Solution>

Specified by:

compare in interface DominanceComparator

Parameters:

solution1 - the first solution

solution2 - the second solution

Returns:

-1 if solution1 dominates solution2, 1 if solution2 dominates solution1, and 0 if the solutions are non-dominated

calculateFitness

public double calculateFitness(Solution solution)

Description copied from interface: AggregateObjectiveComparator

Calculates the aggregate (fitness) value of the solution using this aggregate function.

Specified by:

calculateFitness in interface AggregateObjectiveComparator

Parameters:

solution - the solution

Returns:

the aggregate value (smaller is better)

calculateFitness

public static double calculateFitness(Solution solution,
 double[] weights,
 double q)

Computes the vector angle distance scaling aggregate fitness of the solution. One weight should be given for each objective.

Parameters:

solution - the solution

weights - the weight vector

q - factor for scaling the effects of the angle

Returns:

the fitness, where smaller values are preferred