001 /*
002 * Java Genetic Algorithm Library (jenetics-5.2.0).
003 * Copyright (c) 2007-2020 Franz Wilhelmstötter
004 *
005 * Licensed under the Apache License, Version 2.0 (the "License");
006 * you may not use this file except in compliance with the License.
007 * You may obtain a copy of the License at
008 *
009 * http://www.apache.org/licenses/LICENSE-2.0
010 *
011 * Unless required by applicable law or agreed to in writing, software
012 * distributed under the License is distributed on an "AS IS" BASIS,
013 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
014 * See the License for the specific language governing permissions and
015 * limitations under the License.
016 *
017 * Author:
018 * Franz Wilhelmstötter (franz.wilhelmstoetter@gmail.com)
019 */
020 package io.jenetics;
021
022 import static java.lang.String.format;
023 import static io.jenetics.internal.math.Combinatorics.subset;
024 import static io.jenetics.internal.math.Randoms.indexes;
025
026 import java.util.Random;
027
028 import io.jenetics.util.MSeq;
029 import io.jenetics.util.RandomRegistry;
030 import io.jenetics.util.Seq;
031
032 /**
033 * <p>
034 * An enhanced genetic algorithm (EGA) combine elements of existing solutions in
035 * order to create a new solution, with some of the properties of each parent.
036 * Recombination creates a new chromosome by combining parts of two (or more)
037 * parent chromosomes. This combination of chromosomes can be made by selecting
038 * one or more crossover points, splitting these chromosomes on the selected
039 * points, and merge those portions of different chromosomes to form new ones.
040 * </p>
041 * <p>
042 * The recombination probability <i>P(r)</i> determines the probability that a
043 * given individual (genotype, not gene) of a population is selected for
044 * recombination. The (<i>mean</i>) number of changed individuals depend on the
045 * concrete implementation and can be vary from
046 * <i>P(r)</i>·<i>N<sub>G</sub></i> to
047 * <i>P(r)</i>·<i>N<sub>G</sub></i>·<i>O<sub>R</sub></i>, where
048 * <i>O<sub>R</sub></i> is the order of the recombination, which is the number
049 * of individuals involved int the {@link #recombine} method.
050 * </p>
051 *
052 * @author <a href="mailto:franz.wilhelmstoetter@gmail.com">Franz Wilhelmstötter</a>
053 * @since 1.0
054 * @version 5.2
055 */
056 public abstract class Recombinator<
057 G extends Gene<?, G>,
058 C extends Comparable<? super C>
059 >
060 extends AbstractAlterer<G, C>
061 {
062
063 private final int _order;
064
065 /**
066 * Constructs an alterer with a given recombination probability.
067 *
068 * @param probability The recombination probability.
069 * @param order the number of individuals involved in the
070 * {@link #recombine(MSeq, int[], long)} step
071 * @throws IllegalArgumentException if the {@code probability} is not in the
072 * valid range of {@code [0, 1]} or the given {@code order} is
073 * smaller than two.
074 */
075 protected Recombinator(final double probability, final int order) {
076 super(probability);
077 if (order < 2) {
078 throw new IllegalArgumentException(format(
079 "Order must be greater than one, but was %d.", order
080 ));
081 }
082 _order = order;
083 }
084
085 /**
086 * Return the number of individuals involved in the
087 * {@link #recombine(MSeq, int[], long)} step.
088 *
089 * @return the number of individuals involved in the recombination step.
090 */
091 public int order() {
092 return _order;
093 }
094
095 /**
096 * Return the number of individuals involved in the
097 * {@link #recombine(MSeq, int[], long)} step.
098 *
099 * @return the number of individuals involved in the recombination step.
100 * @deprecated Use {@link #order()} instead
101 */
102 @Deprecated
103 public int getOrder() {
104 return _order;
105 }
106
107 @Override
108 public final AltererResult<G, C> alter(
109 final Seq<Phenotype<G, C>> population,
110 final long generation
111 ) {
112 final AltererResult<G, C> result;
113 if (population.size() >= 2) {
114 final Random random = RandomRegistry.random();
115 final int order = Math.min(_order, population.size());
116
117 final MSeq<Phenotype<G, C>> pop = MSeq.of(population);
118 final int count = indexes(random, population.size(), _probability)
119 .mapToObj(i -> individuals(i, population.size(), order, random))
120 .mapToInt(ind -> recombine(pop, ind, generation))
121 .sum();
122
123 result = AltererResult.of(pop.toISeq(), count);
124 } else {
125 result = AltererResult.of(population.asISeq());
126 }
127
128 return result;
129 }
130
131 static int[] individuals(
132 final int index,
133 final int size,
134 final int order,
135 final Random random
136 ) {
137 final int[] ind = subset(size, order, random);
138
139 // Find the correct slot for the "master" individual.
140 // This prevents duplicate index entries.
141 int i = 0;
142 while (ind[i] < index && i < ind.length - 1) {
143 ++i;
144 }
145 ind[i] = index;
146
147 return ind;
148 }
149
150 /**
151 * Recombination template method. This method is called 0 to n times. It is
152 * guaranteed that this method is only called by one thread.
153 *
154 * @param population the population to recombine
155 * @param individuals the array with the indexes of the individuals which
156 * are involved in the <i>recombination</i> step. The length of the
157 * array is {@link #order()}. The first individual is the
158 * <i>primary</i> individual.
159 * @param generation the current generation.
160 * @return the number of genes that has been altered.
161 */
162 protected abstract int recombine(
163 final MSeq<Phenotype<G, C>> population,
164 final int[] individuals,
165 final long generation
166 );
167
168 }
|