Evolutionpy

Work in Progress [WIP]

Just another evolution based optimizer built using python.

NOTE: Created to learn how python package is build.

Evolution Algorithm We handle currently

• Genetic Algorithm [Not Completely] [For 1st Release]

Minimal Example

Other Examples can be found on examples/

import numpy as np

from evolutionpy.modules.crossover import SinglePointCrossover
from evolutionpy.modules.ga import BaseGA
from evolutionpy.modules.mutation import RandomMutation
from evolutionpy.modules.selection import SteadyStateSelection
from evolutionpy.structures import GeneSpace


# Given the following function:
# f(x,y) = xy - x^2 - y^2 - 2x - 2y + 4
# Here, f(x,y) attends maximum value f_max=8 at the point (x,y) = (-2,-2)
def fitness_func(individual: np.ndarray, **kwargs) -> float:
  """Fitness function for above equations."""
  x = individual[0]
  y = individual[1]
  return x * y - x**2 - y**2 - 2 * x - 2 * y + 4


class CustomGA(BaseGA):
  """Custom GA for the above problem."""

  def set_operators(self):
    """Set up operators."""
    # self.default_operator()
    self.selection = SteadyStateSelection(self.population, self.population_size)
    self.crossover_probability = 0.3
    self.crossover = SinglePointCrossover(
      crossover_probability=self.crossover_probability
    )
    self.mutation_rate = 0.1
    self.mutation = RandomMutation(mutation_rate=self.mutation_rate)


if __name__ == "__main__":
  num_generations = 100
  population_size = 100

  bounds = [[-10, 10], [-10, 10]]

  gene_space = GeneSpace(bounds)

  custom_ga = CustomGA(
    gene_space=gene_space,
    fitness_func=fitness_func,
    num_generations=num_generations,
    population_size=population_size,
  )
  # Running the GA to optimize the parameters of the function.
  custom_ga.run()

TODO / Future Plans

• Docs for getting started
• More operators
• Integrate with sklearn for hyperparameter optimization

Credits and Package to checkout

• Motivated from PyGAD : Genetic Algorithm in Python
• https://github.com/gugarosa/opytimizer
• https://github.com/ljvmiranda921/pyswarms

Contribution Guidelines

• Currently not looking for contributions