001 /*
002 * Java Genetic Algorithm Library (jenetics-1.5.0).
003 * Copyright (c) 2007-2013 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@gmx.at)
019 */
020 package org.jenetics;
021
022 import static java.lang.Math.pow;
023 import static java.lang.String.format;
024 import static org.jenetics.util.object.hashCodeOf;
025
026 import java.util.concurrent.atomic.AtomicInteger;
027
028 import org.jenetics.util.IndexStream;
029 import org.jenetics.util.MSeq;
030
031
032 /**
033 * This class is for mutating a chromosomes of an given population. There are
034 * two distinct roles mutation plays
035 * <ul>
036 * <li>Exploring the search space. By making small moves mutation allows a
037 * population to explore the search space. This exploration is often slow
038 * compared to crossover, but in problems where crossover is disruptive this
039 * can be an important way to explore the landscape.
040 * </li>
041 * <li>Maintaining diversity. Mutation prevents a population from
042 * correlating. Even if most of the search is being performed by crossover,
043 * mutation can be vital to provide the diversity which crossover needs.
044 * </li>
045 * </ul>
046 *
047 * <p>
048 * The mutation probability is the parameter that must be optimized. The optimal
049 * value of the mutation rate depends on the role mutation plays. If mutation is
050 * the only source of exploration (if there is no crossover) then the mutation
051 * rate should be set so that a reasonable neighborhood of solutions is explored.
052 * </p>
053 * The mutation probability <i>P(m)</i> is the probability that a specific gene
054 * over the whole population is mutated. The number of available genes of an
055 * population is
056 * <p>
057 * <img src="doc-files/mutator-N_G.gif" alt="N_P N_{g}=N_P \sum_{i=0}^{N_{G}-1}N_{C[i]}" />
058 * </p>
059 * where <i>N<sub>P</sub></i> is the population size, <i>N<sub>g</sub></i> the
060 * number of genes of a genotype. So the (average) number of genes
061 * mutated by the mutation is
062 * <p>
063 * <img src="doc-files/mutator-mean_m.gif" alt="\hat{\mu}=N_{P}N_{g}\cdot P(m)" />
064 * </p>
065 *
066 * @author <a href="mailto:franz.wilhelmstoetter@gmx.at">Franz Wilhelmstötter</a>
067 * @since 1.0
068 * @version 1.0 — <em>$Date: 2013-12-02 $</em>
069 */
070 public class Mutator<G extends Gene<?, G>> extends AbstractAlterer<G> {
071
072 /**
073 * Construct a Mutation object which a given mutation probability.
074 *
075 * @param probability Mutation probability. The given probability is
076 * divided by the number of chromosomes of the genotype to form
077 * the concrete mutation probability.
078 * @throws IllegalArgumentException if the {@code probability} is not in the
079 * valid range of {@code [0, 1]}..
080 */
081 public Mutator(final double probability) {
082 super(probability);
083 }
084
085 /**
086 * Default constructor, with probability = 0.01.
087 */
088 public Mutator() {
089 this(0.01);
090 }
091
092 /**
093 * Concrete implementation of the alter method.
094 */
095 @Override
096 public <C extends Comparable<? super C>> int alter(
097 final Population<G, C> population,
098 final int generation
099 ) {
100 assert(population != null) : "Not null is guaranteed from base class.";
101
102 final double p = pow(_probability, 1.0/3.0);
103 final AtomicInteger alterations = new AtomicInteger(0);
104
105 final IndexStream stream = IndexStream.Random(population.size(), p);
106 for (int i = stream.next(); i != -1; i = stream.next()) {
107 final Phenotype<G, C> pt = population.get(i);
108
109 final Genotype<G> gt = pt.getGenotype();
110 final Genotype<G> mgt = mutate(gt, p, alterations);
111
112 final Phenotype<G, C> mpt = pt.newInstance(mgt, generation);
113 population.set(i, mpt);
114 }
115
116 return alterations.get();
117 }
118
119 private Genotype<G> mutate(
120 final Genotype<G> genotype,
121 final double p,
122 final AtomicInteger alterations
123 ) {
124 Genotype<G> gt = genotype;
125
126 final IndexStream stream = IndexStream.Random(genotype.length(), p);
127 final int start = stream.next();
128
129 if (start != -1) {
130 final MSeq<Chromosome<G>> chromosomes = genotype.toSeq().copy();
131
132 for (int i = start; i != -1; i = stream.next()) {
133 final Chromosome<G> chromosome = chromosomes.get(i);
134 final MSeq<G> genes = chromosome.toSeq().copy();
135
136 final int mutations = mutate(genes, p);
137 if (mutations > 0) {
138 alterations.addAndGet(mutations);
139 chromosomes.set(i, chromosome.newInstance(genes.toISeq()));
140 }
141 }
142
143 gt = genotype.newInstance(chromosomes.toISeq());
144 }
145
146 return gt;
147 }
148
149 /**
150 * <p>
151 * Template method which gives an (re)implementation of the mutation class
152 * the possibility to perform its own mutation operation, based on a
153 * writable gene array and the gene mutation probability <i>p</i>.
154 * </p>
155 * This implementation, for example, does it in this way:
156 * [code]
157 * protected int mutate(final MSeq〈G〉 genes, final double p) {
158 * final IndexStream stream = IndexStream.Random(genes.length(), p);
159 *
160 * int alterations = 0;
161 * for (int i = stream.next(); i != -1; i = stream.next()) {
162 * genes.set(i, genes.get(i).newInstance());
163 * ++alterations;
164 * }
165 * return alterations;
166 * }
167 * [/code]
168 *
169 * @param genes the genes to mutate.
170 * @param p the gene mutation probability.
171 */
172 protected int mutate(final MSeq<G> genes, final double p) {
173 final IndexStream stream = IndexStream.Random(genes.length(), p);
174
175 int alterations = 0;
176 for (int i = stream.next(); i != -1; i = stream.next()) {
177 genes.set(i, genes.get(i).newInstance());
178 ++alterations;
179 }
180
181 return alterations;
182 }
183
184 @Override
185 public int hashCode() {
186 return hashCodeOf(getClass()).and(super.hashCode()).value();
187 }
188
189 @Override
190 public boolean equals(final Object obj) {
191 return obj == this || obj instanceof Mutator<?>;
192 }
193
194 @Override
195 public String toString() {
196 return format("%s[p=%f]", getClass().getSimpleName(), _probability);
197 }
198
199 }
200
201
202
|