Genetic Algorithms Library
AForge.NET framework provides genetic algorithms library, which contains set of classes
providing functionality allowing to solve many different problems with the help of evolutionary
computations based on Genetic Algorithms (GA), Genetic Programming (GP) and Gene Expression
Programming (GEP).
The library consists of 3 main sets of classes:
- Collection of different chromosomes' classes implementing
IChromosome
interface suiting different common problems. These classes implement such operators like
crossover and mutation, which are used for driving search through solutions' space by
creating new chromosomes representing new solutions.
- Collection of fitness functions' classes implementing
IFitnessFunction
interface, which are used to evaluate chromosomes by calculating their fitness values
for different tasks;
- Collection of selection algorithms' classes implementing
ISelectionMethod
interface, which are used to perform selection of population members (chromosomes).
The main class of the library is
Population
class, which organizes the work of genetic algorithm (GA/GP/GEP) creating initial population
of random members, creating new members with the help of crossover and mutations operators,
calculating fitness values of new members and performing selection of members to keep basing on
members' usefulness (fitness). Creating population object it is required to specify which chromosomes,
fitness function and selection algorithm to use. Once this is done, the population object is ready
to be used for searching of problem's solution:
// create genetic population
Population population = new Population( 40,
baseChromosome,
fitnessFunction,
selectionAlgorithm );
while ( true )
{
// run one epoch of the population
population.RunEpoch( );
// check current best fitness
if ( population.FitnessMax > limit )
{
// ...
}
}
As an example, below is small a sample code, which demonstrates usage of
Population class
for solving quite simple problem - searching function's maximum value.
// define optimization function
public class UserFunction : OptimizationFunction1D
{
public UserFunction( ) :
base( new DoubleRange( 0, 255 ) ) { }
public override double OptimizationFunction( double x )
{
return Math.Cos( x / 23 ) * Math.Sin( x / 50 ) + 2;
}
}
...
// create genetic population
Population population = new Population( 40,
new BinaryChromosome( 32 ),
new UserFunction( ),
new EliteSelection( ) );
// run one epoch of the population
population.RunEpoch( );
The Population class
also supports migration of members
from one population to another, which allows to exchange by good solutions between populations.
Also this feature allows to run several populations simultaneously in different threads (distributing computations on multiple cores/CPUs
if they are available) and exchange by good solutions from time to time bringing "fresh blood" to populations.
For additional information check
genetic algorithms' samples
provided with AForge.NET framework.
|
