pyalgen 0.0.6

Minimal python library for genetic algorithm

pyalgen - A minimal library for genetic algorithm in python

Install

pip3 install pyalgen

How to use

1. Import the package

import pyalgen # import pyalgen
from pyalgen import TestFunctions as tf # test functions to optimize

2. Define population

pop = pyalgen.Population(low=-10, high=10, dtype='float', dist='uniform')
# the variable in the population is of type float and the 
# values are taken from a uniform distribution in (low, high)
population = pop(pop_size=1000, variables=2)
# variables is the number of variables to optimize. We are optimizing the 
# matyas function which has two variables
# search space ranges from -10 to 10
# pop_size is the population size of each generation

Matyas function

3. Select type of selection, crossover and mutation strategies

selection = pyalgen.Selection.tournament
crossover = pyalgen.Crossover.onepoint
mutation = pyalgen.Mutation.randompoint

4. Instantiate Genetic Algorithm object with defined variables

ga = pyalgen.GeneticAlgorithm(population, tf.matyas, selection, crossover, mutation)

5. Run the algorithm

iterations, objective, pop = ga.forward(iterations=200)
# iterations is the number of generations to run for
print(f'objective: {objective.min()}, solution: {pop[objective.argmin()]}, generation: {iterations}')
# print the minimum objective and the chromosome in population which 
# given minimum objective 

6. Check the result

100%|██████████████████████████████████████████████████████████| 1000/1000 [00:03<00:00, 262.55it/s]
objective: 7.719286052427051e-07, solution: [-0.00447918 -0.00410235], generation: 1000
# global minimum of matyas is at f(0, 0) = 0
# our algorithm gives minimum, f(-0.004, -0.004) = 7.7e-07
# which is pretty close 

Results can be improved by tweaking the parameters

Testing the algorithm on custom function

Let's solve the equation,

Complete code

import pyalgen

pop = pyalgen.Population(1, 30, unique=True, dtype='int')
population = pop(1000, 4) # here we generate intergers in range[1, 30)
# for population

selection = pyalgen.Selection.tournament
crossover = pyalgen.Crossover.onepoint
mutation = pyalgen.Mutation.randompoint

def obj(a, b, c, d): # objective function
    return a + 2*b + 3*c + 4*d - 30 

ga = pyalgen.GeneticAlgorithm(population, obj, selection, crossover, mutation)


iterations, objective, pop = ga.forward(iterations=50)

if iterations == 1000:
    print(f'objective: {objective.min()}, \
        solution: {pop[objective.argmin()]}, generation: {iterations}')   
else:
    print(f'objective: {objective[objective == 0][0]},\
         solution: {pop[objective == 0][0]}, generation: {iterations}')   

GeneticAlgorithm breaks the computation, if any of the chromosome reached out objective, i.e, 0

Result

 40%|████████████████████████▊                                     | 20/50 [00:00<00:00, 509.30it/s]
objective: 0, solution: [4 6 2 2], generation: 20

The algorithm reached a solution during generation: 11 Solution: a = 11, b = 4, c = 1, d = 2