org.uncommons.watchmaker.framework.operators
Class ListOrderMutation<T>

java.lang.Object
  org.uncommons.watchmaker.framework.operators.ListOrderMutation<T>

Type Parameters:

T - The component type of the lists that are mutated.

All Implemented Interfaces:

EvolutionaryOperator<List<T>>

public class ListOrderMutation<T>
extends Object
implements EvolutionaryOperator<List<T>>

A special mutation implementation that instead of changing the genes of the candidate, re-orders them. A single mutation involves swapping a random element in the list with the element immediately after it. This operation can either apply a fixed number of mutations to each candidate or it can draw values from a random sequence, typically a poisson distribution (see PoissonGenerator), to determine how many mutations to apply.

Author:

Daniel Dyer

Constructor Summary

ListOrderMutation()
Default is one mutation per candidate.

ListOrderMutation(int mutationCount, int mutationAmount)

ListOrderMutation(NumberGenerator<Integer> mutationCount, NumberGenerator<Integer> mutationAmount)
Typically the mutation count will be from a Poisson distribution.

Method Summary

List<List<T>>

apply(List<List<T>> selectedCandidates, Random rng) Apply the operation to each entry in the list of selected candidates.

Methods inherited from class java.lang.Object

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

Constructor Detail

ListOrderMutation

public ListOrderMutation()

Default is one mutation per candidate.

ListOrderMutation

public ListOrderMutation(int mutationCount,
                         int mutationAmount)

Parameters:

mutationCount - The constant number of mutations to apply to each individual in the population.

mutationAmount - The constant number of positions by which a list element will be displaced as a result of mutation.

ListOrderMutation

public ListOrderMutation(NumberGenerator<Integer> mutationCount,
                         NumberGenerator<Integer> mutationAmount)

Typically the mutation count will be from a Poisson distribution. The mutation amount can be from any discrete probability distribution and can include negative values.

Parameters:

mutationCount - A random variable that provides a number of mutations that will be applied to each individual.

mutationAmount - A random variable that provides a number of positions by which to displace an element when mutating.

Method Detail

apply

public List<List<T>> apply(List<List<T>> selectedCandidates,
                           Random rng)

Description copied from interface: EvolutionaryOperator

Apply the operation to each entry in the list of selected candidates. It is important to note that this method operates on the list of candidates returned by the selection strategy and not on the current population. Each entry in the list (not each individual - the list may contain the same individual more than once) must be operated on exactly once.

Implementing classes should not assume any particular ordering (or lack of ordering) for the selection. If ordering or shuffling is required, it should be performed by the implementing class. The implementation should not re-order the list provided but instead should make a copy of the list and re-order that. The ordering of the selection should be totally irrelevant for operators that process each candidate in isolation, such as mutation. It should only be an issue for operators, such as cross-over, that deal with multiple candidates in a single operation.

The operator should not modify any of the candidates passed in. Instead it should return a list that contains evolved copies of those candidates (umodified candidates can be included in the results without having to be copied).

Specified by:

apply in interface EvolutionaryOperator<List<T>>

Parameters:

selectedCandidates - The individuals to evolve.

rng - A source of randomness for stochastic operators (most operators will be stochastic).

Returns:

The evolved individuals.