Jenetics is an Genetic Algorithm, respectively an Evolutionary Algorithm, library written in Java. It is designed with a clear separation of the several algorithm concepts, e. g. Gene, Chromosome, Genotype, Phenotype, Population and fitness Function. Jenetics allows you to minimize or maximize the given fitness function without tweaking it.
- JDK 7: The
JAVA_HOMEvariable must be set to your java installation directory. - Gradle 1.10: Gradle is used for building the library. (Gradle is download automatically, if you are using the Gradle Wrapper script
gradlew, located in the base directory, for building the library.)
- TestNG 8.8: Jenetics uses TestNG framework for unit tests.
- Apache Commons Math 3.2: Library is used for testing statistical accumulators.
- JRE 7: Java runtime version 7 is needed for using the library, respectively for running the examples.
- Sourceforge: https://sourceforge.net/projects/jenetics/files/latest/download
- Bitbucket: https://bitbucket.org/fwilhelm/jenetics/downloads
- Maven:
org.bitbucket.fwilhelm:org.jenetics:2.0.0on Maven Central
For building the Jenetics library from source, download the most recent, stable package version from Sourceforge or Bitbucket and extract it to some build directory.
$ unzip jenetics-<version>.zip -d <builddir>
<version> denotes the actual Jenetics version and <builddir> the actual build directory. Alternatively you can check out the latest-unstable-version from the Mercurial default branch.
$ hg clone https://bitbucket.org/fwilhelm/jenetics <builddir>
# or
$ hg clone http://hg.code.sf.net/p/jenetics/main <builddir>
# or
$ git clone https://github.com/jenetics/jenetics.git <builddir>
Jenetics uses Gradle as build system and organizes the source into sub-projects (modules). Each sub-project is located in it’s own sub-directory:
- org.jenetics: This project contains the source code and tests for the Jenetics core-module.
- org.jenetics.example: This project contains example code for the core-module.
- org.jenetics.doc: Contains the code of the web-site and the manual.
For building the library change into the <builddir> directory (or one of the module directory) and call one of the available tasks:
- compileJava: Compiles the Jenetics sources and copies the class files to the
<builddir>/<module-dir>/build/classes/maindirectory. - test: Compiles and executes the unit tests. The test results are printed onto the console and a test-report, created by TestNG, is written to
<builddir>/<module-dir>directory. - javadoc: Generates the API documentation. The Javadoc is stored in the
<builddir>/<module-dir>/build/docsdirectory - jar: Compiles the sources and creates the JAR files. The artifacts are copied to the
<builddir>/<module-dir>/build/libsdirectory. - packaging: Compiles the sources of all modules, creates the JAR files and the Javadoc and creates a complete library package--the same which you can download from the home page. The build artifacts are copied into the
<builddir>/build/package/jenetics-<version>directory. - clean: Deletes the
<builddir>/build/*directories and removes all generated artifacts.
For packaging (building) the source call
$ cd <build-dir>
$ ./gradlew packaging
IDE Integration
Gradle has tasks which creates the project file for Eclipse and IntelliJ IDEA. Call
$ ./gradlew [eclipse|idea]
for creating the project files for Eclipse or IntelliJ, respectively.
Ones counting is one of the simplest model-problem and consists of a binary chromosome. The fitness of a Genotype is proportional to the number of ones. The fitness Function looks like this:
import org.jenetics.BitChromosome;
import org.jenetics.BitGene;
import org.jenetics.GeneticAlgorithm;
import org.jenetics.Genotype;
import org.jenetics.Mutator;
import org.jenetics.NumberStatistics;
import org.jenetics.Optimize;
import org.jenetics.RouletteWheelSelector;
import org.jenetics.SinglePointCrossover;
import org.jenetics.util.Factory;
import org.jenetics.util.Function;
final class OneCounter
implements Function<Genotype<BitGene>, Integer>
{
@Override
public Integer apply(final Genotype<BitGene> genotype) {
return ((BitChromosome)genotype.getChromosome()).bitCount();
}
}
public class OnesCounting {
public static void main(String[] args) {
Factory<Genotype<BitGene>> gtf = Genotype.of(
BitChromosome.of(20, 0.15)
);
Function<Genotype<BitGene>, Integer> ff = new OneCounter();
GeneticAlgorithm<BitGene, Integer> ga =
new GeneticAlgorithm<>(
gtf, ff, Optimize.MAXIMUM
);
ga.setStatisticsCalculator(
new NumberStatistics.Calculator<BitGene, Integer>()
);
ga.setPopulationSize(500);
ga.setSelectors(
new RouletteWheelSelector<BitGene, Integer>()
);
ga.setAlterers(
new Mutator<BitGene>(0.55),
new SinglePointCrossover<BitGene>(0.06)
);
ga.setup();
ga.evolve(100);
System.out.println(ga.getBestStatistics());
System.out.println(ga.getBestPhenotype());
}
}
The genotype in this example consists of one BitChromosome with a ones probability of 0.15. The altering of the offspring population is performed by mutation, with mutation probability of 0.55, and then by a single-point crossover, with crossover probability of 0.06. After creating the initial population, with the ga.setup() call, 100 generations are evolved. The tournament selector is used for both, the offspring- and the survivor selection-this is the default selector.
+---------------------------------------------------------+
| Population Statistics |
+---------------------------------------------------------+
| Age mean: 1.14800000000 |
| Age variance: 2.88386372745 |
| Samples: 500 |
| Best fitness: 19 |
| Worst fitness: 4 |
+---------------------------------------------------------+
+---------------------------------------------------------+
| Fitness Statistics |
+---------------------------------------------------------+
| Fitness mean: 11.00400000000 |
| Fitness variance: 6.28856112224 |
| Fitness error of mean: 0.49211384049 |
+---------------------------------------------------------+
[00001111|11111111|11111011] --> 19
The given example will print the overall timing statistics onto the console.
In the knapsack problem a set of items, together with their size and value, is given. The task is to select a disjoint subset so that the total size does not exeed the knapsacks size. For the 0/1 knapsack problem we define a BitChromosome, one bit for each item. If the ith BitGene is set to one the ith item is selected.
import static org.jenetics.util.math.random.nextDouble;
import java.util.Random;
import org.jenetics.BitChromosome;
import org.jenetics.BitGene;
import org.jenetics.Chromosome;
import org.jenetics.GeneticAlgorithm;
import org.jenetics.Genotype;
import org.jenetics.Mutator;
import org.jenetics.NumberStatistics;
import org.jenetics.RouletteWheelSelector;
import org.jenetics.SinglePointCrossover;
import org.jenetics.TournamentSelector;
import org.jenetics.util.Factory;
import org.jenetics.util.Function;
import org.jenetics.util.LCG64ShiftRandom;
import org.jenetics.util.RandomRegistry;
import org.jenetics.util.Scoped;
final class Item {
public final double size;
public final double value;
Item(final double size, final double value) {
this.size = size;
this.value = value;
}
}
final class KnapsackFunction
implements Function<Genotype<BitGene>, Double>
{
private final Item[] items;
private final double size;
public KnapsackFunction(final Item[] items, double size) {
this.items = items;
this.size = size;
}
@Override
public Double apply(final Genotype<BitGene> genotype) {
final Chromosome<BitGene> ch = genotype.getChromosome();
double size = 0;
double value = 0;
for (int i = 0, n = ch.length(); i < n; ++i) {
if (ch.getGene(i).getBit()) {
size += items[i].size;
value += items[i].value;
}
}
return size <= this.size ? value : 0;
}
}
public class Knapsack {
private static KnapsackFunction FF(final int n, final double size) {
final Item[] items = new Item[n];
try (Scoped<? extends Random> random =
RandomRegistry.scope(new LCG64ShiftRandom(123)))
{
for (int i = 0; i < items.length; ++i) {
items[i] = new Item(
nextDouble(random.get(), 0, 100),
nextDouble(random.get(), 0, 100)
);
}
}
return new KnapsackFunction(items, size);
}
public static void main(String[] args) throws Exception {
final int nitems = 15;
final double kssize = nitems*100.0/3.0;
final KnapsackFunction ff = FF(nitems, kssize);
final Factory<Genotype<BitGene>> genotype = Genotype.of(
BitChromosome.of(nitems, 0.5)
);
final GeneticAlgorithm<BitGene, Double> ga = new GeneticAlgorithm<>(
genotype, ff
);
ga.setPopulationSize(500);
ga.setStatisticsCalculator(
new NumberStatistics.Calculator<BitGene, Double>()
);
ga.setSurvivorSelector(
new TournamentSelector<BitGene, Double>(5)
);
ga.setOffspringSelector(
new RouletteWheelSelector<BitGene, Double>()
);
ga.setAlterers(
new Mutator<BitGene>(0.115),
new SinglePointCrossover<BitGene>(0.16)
);
ga.setup();
ga.evolve(100);
System.out.println(ga.getBestStatistics());
System.out.println(ga.getBestPhenotype());
}
}
The console out put for the Knapsack GA will look like the listing beneath.
+---------------------------------------------------------+
| Population Statistics |
+---------------------------------------------------------+
| Age mean: 2.38000000000 |
| Age variance: 6.22004008016 |
| Samples: 500 |
| Best fitness: 643.239770840163 |
| Worst fitness: 0.0 |
+---------------------------------------------------------+
+---------------------------------------------------------+
| Fitness Statistics |
+---------------------------------------------------------+
| Fitness mean: 525.20849288163 |
| Fitness variance: 20182.61311489490 |
| Fitness error of mean: 23.48803784887 |
+---------------------------------------------------------+
[01101111|01011111] --> 643.239770840163
The Traveling Salesman problem is a very good example which shows you how to solve combinatorial problems with an GA. Jenetics contains several classes which will work very well with this kind of problems. Wrapping the base type into an EnumGene is the first thing to do. In our example, every city has an unique number, that means we are wrapping an Integer into an EnumGene. Creating a genotype for integer values is very easy with the factory method of the PermutationChromosome. For other data types you have to use one of the constructors of the permutation chromosome. As alterers, we are using a swap-mutator and a partially-matched crossover. These alterers guarantees that no invalid solutions are created—every city exists exactly once in the altered chromosomes.
import static java.lang.Math.PI;
import static java.lang.Math.abs;
import static java.lang.Math.sin;
import java.io.Serializable;
import org.jenetics.Chromosome;
import org.jenetics.EnumGene;
import org.jenetics.GeneticAlgorithm;
import org.jenetics.Genotype;
import org.jenetics.NumberStatistics.Calculator;
import org.jenetics.Optimize;
import org.jenetics.PartiallyMatchedCrossover;
import org.jenetics.PermutationChromosome;
import org.jenetics.SwapMutator;
import org.jenetics.util.Factory;
import org.jenetics.util.Function;
public class TravelingSalesman {
private static class FF
implements Function<Genotype<EnumGene<Integer>>, Double>,
Serializable
{
private static final long serialVersionUID = 1L;
private final double[][] adjacence;
public FF(final double[][] adjacence) {
this.adjacence = adjacence;
}
@Override
public Double apply(final Genotype<EnumGene<Integer>> genotype) {
final Chromosome<EnumGene<Integer>> path = genotype.getChromosome();
double length = 0.0;
for (int i = 0, n = path.length(); i < n; ++i) {
final int from = path.getGene(i).getAllele();
final int to = path.getGene((i + 1)%n).getAllele();
length += adjacence[from][to];
}
return length;
}
@Override
public String toString() {
return "Point distance";
}
}
public static void main(String[] args) {
final int stops = 20;
final Function<Genotype<EnumGene<Integer>>, Double> ff = new FF(adjacencyMatrix(stops));
final Factory<Genotype<EnumGene<Integer>>> gtf = Genotype.of(
PermutationChromosome.ofInteger(stops)
);
final GeneticAlgorithm<EnumGene<Integer>, Double>
ga = new GeneticAlgorithm<>(gtf, ff, Optimize.MINIMUM);
ga.setStatisticsCalculator(
new Calculator<EnumGene<Integer>, Double>()
);
ga.setPopulationSize(500);
ga.setAlterers(
new SwapMutator<EnumGene<Integer>>(0.2),
new PartiallyMatchedCrossover<Integer>(0.3)
);
ga.setup();
ga.evolve(100);
System.out.println(ga.getBestStatistics());
System.out.println(ga.getBestPhenotype());
}
/**
* All points in the created adjacency matrix lie on a circle. So it is easy
* to check the quality of the solution found by the GA.
*/
private static double[][] adjacencyMatrix(int stops) {
double[][] matrix = new double[stops][stops];
for (int i = 0; i < stops; ++i) {
for (int j = 0; j < stops; ++j) {
matrix[i][j] = chord(stops, abs(i - j), RADIUS);
}
}
return matrix;
}
private static double chord(int stops, int i, double r) {
return 2.0*r*abs(sin((PI*i)/stops));
}
private static double RADIUS = 10.0;
}
The listing above shows the output generated by our example. The last line represents the phenotype of the best solution found by the GA, which represents the traveling path. As you can see, the GA has found the shortest path, in reverse order.
+---------------------------------------------------------+
| Population Statistics |
+---------------------------------------------------------+
| Age mean: 1.67600000000 |
| Age variance: 4.17537474950 |
| Samples: 500 |
| Best fitness: 104.39898868995472 |
| Worst fitness: 310.43291286365866 |
+---------------------------------------------------------+
+---------------------------------------------------------+
| Fitness Statistics |
+---------------------------------------------------------+
| Fitness mean: 132.54050652459 |
| Fitness variance: 3230.15643373580 |
| Fitness error of mean: 5.92739164722 |
+---------------------------------------------------------+
[1|0|19|16|15|5|6|7|8|9|10|11|12|13|14|17|18|4|3|2] --> 104.39898868995472
Beside the Java coding standards as given in http://www.oracle.com/technetwork/java/javase/documentation/codeconvtoc-136057.html the following extensions are used.
- All non-constant variables members start with underscore.
- Variable name for arrays or collections are plural.
- All helper classes which only contains static methods are lower-case. This indicates that the given class can not be used as type, because no instance can be created.
The library is licensed under the Apache License, Version 2.0.
Copyright 2007-2014 Franz Wilhelmstötter
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
- Add IntegerGene/Chromosome classes.
- Remove all deprecated classes and methods.
- Remove dependency to the JScience library.
- All concurrency classes are now internal.
GeneticAlgorithmclass takes anExecutoras additional parameter, which is used for parallelizable code. - Library can now be downloaded via the maven central repository:
org.bitbucket.fwilhelm:org.jenetics:2.0.0
- Preparation work for removing the dependency to the JScience library.
- Add Double/Long Gene/Chromosome as a replacement for Float64/Integer64 Gene/Chromosome.
- Add JAXB XML serialization as a replacement of the Javolution XML marshalling.
- Streamlining of the existing API: Marking inconsistent methods/classes as deprecated.