Usage

In this section of the documentation an small example of how to use the genetic algorithm (GA) module is given.

For the example we are going to use as the optimization objective the Ackley function (ackley).

First we need to import the GA module and the Ackley function:

from mango.benchmark.optimization.ackley import ackley
from mango.models.genetic.population import Population
from mango.models.genetic.config import GeneticBaseConfig

Then we have to define the configuration that we are going to use on a .yml file. We can write the following configuration to a file called ackley.cfg:

[main]
population_size         = 50
max_generations         = 300
optimization_objective  = min
selection               = roulette
crossover               = blend
replacement             = elitist
mutation_control        = static
mutation_base_rate      = 0.2

[individual]
encoding        = real
gene_length     = 100
gene_min_value  = -32.768
gene_max_value  = 32.768

[crossover]
blend_expansion   = 0.5

Then we can read the configuration:

config = GeneticBaseConfig('ackley.cfg')

Now we can create the population:

population = Population(config, ackley)

And we run the GA:

population.run()

Once the model has finished we can get the best individual, its fitness and solution:

best_individual = population.best.individual
best_fitness = best_inidividual.fitness
best_solution = best_individual.genes

As it can be seen running a genetic algorithm to optimize a function is very easy and requires a few lines of code. In order to use any other function we can just have a method that receives the genes (as the ackley function does) or we can use the Problem, creating a subclass and defining the calculate_fitness method.

This would be done as follows:

from mango.models.genetic.problem import Problem

class MyProblem(Problem):
    def calculate_fitness(self, genes):
        # Calculate the fitness of the individual
        # ...
        return fitness
problem = MyProblem()
population = Population(config, problem)
population.run()