geneticd
Simple Genetic Algorithm Library for D.
Instalation
It is recommended to use DUB for including the library to other projects.
Add the following dependency to the package.json file in your project's directory:
{
"name": "your-project-identifier",
"dependencies": {
"geneticd": "~master"
}
}
Status
Library can be used to solve basic problems already. Currently encoding with scalar value gene chromosomes is supported. Common selection and crossover operators are implemented.
Note
I'm not a genetic or evolution algorithms expert nor experienced D programmer (still learning). So any advice is welcome :)
Sample
For advanced samples look to the examples directory in the library.
// Guessing of the array content
alias Chromosome!(ScalarGene!bool) chromoType; // define chromosome type
//init target array
enum size = 20;
bool[] target;
foreach(i; 0..size)
{
target ~= dice(0.5, 0.5) == 1;
}
//create GA configuration
auto conf = new Configuration!chromoType(new chromoType(new ScalarGene!bool(), size)); //config with sample chromosome to init population with
conf.populationSize = 10; //size of the population
//set fitness function
conf.fitnessFunction = simpleFitness!chromoType(delegate (ch)
{
//add to tmp if chromosome and target are same so max fitness = 20
uint tmp;
foreach(i; 0..size)
{
if(target[i] == ch[i]) ++tmp;
}
return cast(double)tmp;
});
//set terminate function
conf.terminateFunction = compositeTerminate( //GA is terminated if one of conditions is met
maxGenerationsTerminate!(100), //limit to max 100 generations
fitnessTerminate!size); //target fitness is size of genomes
//add GA operations
conf.eliteSelectionOperator = eliteSelection!chromoType; // best chromosome allways survives
conf.parentSelectionOperator = weightedRouletteSelection!chromoType(); //select parent chromosomes to crossover and mutate
conf.crossoverOperator = singlePointCrossover!chromoType(); //type of crossover operator
//execute GA
auto ga = new GA!chromoType(conf);
ga.run(); //does return after one of terminate condition is met
writefln("Input: [%s]", target.map!(to!string).joiner(", "));
writefln("Best: [%s]", ga.population.best);TODO
- Scalar gene with boundaries
- Traveling salesman problem example
- Knapsack problem example
- Target number game example
- Saving/loading population to/from file
- Parallel processing (probably chunk the population and fitness chunks in own tasks)
- ??? (fill the bug or pull request)
