Package io.jenetics.ext.moea

Class UFTournamentSelector<G extends Gene<?,G>,C extends Comparable<? super C>>

java.lang.Object

io.jenetics.ext.moea.UFTournamentSelector<G,C>

All Implemented Interfaces:

Selector<G,C>

public class UFTournamentSelector<G extends Gene<?,G>,C extends Comparable<? super C>>
extends Object
implements Selector<G,C>

Unique fitness based tournament selection.

The selection of unique fitnesses lifts the selection bias towards over-represented fitnesses by reducing multiple solutions sharing the same fitness to a single point in the objective space. It is therefore no longer required to assign a crowding distance of zero to individual of equal fitness as the selection operator correctly enforces diversity preservation by picking unique points in the objective space.

Reference: Félix-Antoine Fortin and Marc Parizeau. 2013. Revisiting the NSGA-II crowding-distance computation. In Proceedings of the 15th annual conference on Genetic and evolutionary computation (GECCO '13), Christian Blum (Ed.). ACM, New York, NY, USA, 623-630. DOI= 10.1145/2463372.2463456

Since:

4.1

Version:

4.1

Constructor Summary

Constructors

Constructor	Description
UFTournamentSelector(Comparator<? super C> dominance, ElementComparator<? super C> comparator, ElementDistance<? super C> distance, ToIntFunction<? super C> dimension)	Creates a new UFTournamentSelector with the functions needed for handling the multi-objective result type C.

Method Summary

Modifier and Type	Method	Description
static <G extends Gene<?,G>,T,V extends Vec<T>> UFTournamentSelector<G,V>	ofVec()	Return a new selector for the given result type V.
ISeq<Phenotype<G,C>>	select(Seq<Phenotype<G,C>> population, int count, Optimize opt)

Methods inherited from class java.lang.Object

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

Constructor Detail

UFTournamentSelector

public UFTournamentSelector(Comparator<? super C> dominance,
                            ElementComparator<? super C> comparator,
                            ElementDistance<? super C> distance,
                            ToIntFunction<? super C> dimension)

Creates a new UFTournamentSelector with the functions needed for handling the multi-objective result type C. For the Vec classes, a selector is created like in the following example:

new UFTournamentSelector<>( Vec<T>::dominance, Vec<T>::compare, Vec<T>::distance, Vec<T>::length );

Parameters:

dominance - the pareto dominance comparator

comparator - the vector element comparator

distance - the vector element distance

dimension - the dimensionality of vector type C

See Also:

ofVec()

Method Detail

select

public ISeq<Phenotype<G,C>> select(Seq<Phenotype<G,C>> population,
                                         int count,
                                         Optimize opt)

Specified by:

select in interface Selector<G extends Gene<?,G>,C extends Comparable<? super C>>

ofVec

public static <G extends Gene<?,G>,T,V extends Vec<T>> UFTournamentSelector<G,V> ofVec()

Return a new selector for the given result type V. This method is a shortcut for

new UFTournamentSelector<>( Vec<T>::dominance, Vec<T>::compare, Vec<T>::distance, Vec<T>::length );

Type Parameters:

G - the gene type

T - the array type, e.g. double[]

V - the multi object result type vector

Returns:

a new selector for the given result type V