GANN

GANNlib is a simple Genetic Algorithms and Neural Network library, that provides an easy to use interface for creating GAs and NNs quickly.

Genetic Algorithms Example

Create genetic algorithms quickly and easily, here is an example of the GANNlib using 18 lines.

It currently uses roulette selection (a.k.a fitness proportionate selection) for a selection method and single point crossover as the crossover method. More crossover and selection methods will be added soon.

Python Code Example

from GANNlib import *

def fitnessFunc(self):
    return self.getGene("x")                      #simply returns the 'x' gene for the fitness function, x will be maximised

chrom = Chromosome()
chrom.addGene("x", 9)                             #adds a gene 'x' to the chromosome that is 9 bits in size
chrom.fitnessFunc = fitnessFunc                   #use custom fitness function
pop = Population(20, chrom)                       #create population of 20 of the chrom chromosomes, randomly generated

print("Fittest: "+str(pop.getFittest().fitness))  #print the fittest chromosome in the population
print("Average: "+str(pop.getAverageFitness()))   #print the average fitness across the population

for i in range(100):                              #evolve for 100 generations
    pop.evolve()                                  #evolves the population using GA
    
print("Fittest: "+str(pop.getFittest().fitness))  #print the fittest chromosome again at the end
print("Average: "+str(pop.getAverageFitness()))   #..and also print the average again

Example output

Fittest: 433
Average: 197.05
Fittest: 511
Average: 499.0

511 is the max for the x gene since it's a 9 bit gene. (2^9-1)

Neural networks

In progress...