PyGAD - Genetic Algorithm in Python

3.6.0 verified Fri May 01 auth: no python

PyGAD is a Python library for building genetic algorithms and training machine learning models (Keras & PyTorch). Current version 3.6.0, released March 2025, with a cadence of several minor releases per year. Supports single- and multi-objective optimization, custom operators, and integration with deep learning frameworks.

pip install pygad

Common errors

error TypeError: 'GA' object is not callable ↓

cause In versions <3.0.0, the GA class was instantiated directly; in 3.0.0+, the module import changed.

fix

Use 'import pygad' and then 'ga = pygad.GA(...)'.

error Fitness function must have 2 arguments, but it has 1 ↓

cause After PyGAD 3.0.0, fitness function signature changed to (solution, solution_idx).

fix

Change your fitness function to accept a second argument: def fitness_func(solution, solution_idx):

Warnings

breaking In PyGAD 3.0.0+, the fitness function signature changed: it now receives two arguments (function, solution_idx). The first argument is the solution array, the second is the solution index. ↓

fix Update fitness function from fitness_func(solution) to fitness_func(solution, solution_idx).

gotcha When using 'gene_space' with a step, the gene type is inferred: if step is integer, genes are integers; if step is float, genes are floats. Mixing types can cause unexpected results. ↓

fix Ensure gene_space step matches desired data type.

deprecated The 'delay_after_gen' parameter was removed in v3.4.0. Use 'on_generation' callback to add delay. ↓

fix Replace delay_after_gen with a custom on_generation function containing time.sleep().

Imports

GA
```
import pygad
```
GA is the class name, but you import the module and use pygad.GA.

Quickstart

Minimal example: maximize sum of binary array (genetic algorithm optimization).

import pygad
import numpy as np

# Define fitness function (must accept two arguments: solution and solution_index)
def fitness_func(solution, solution_idx):
    return -np.sum(solution)

# Define gene space
num_genes = 6
gene_space = {'low': 0, 'high': 2, 'step': 1}  # integers 0 or 1

# Create GA instance
ga_instance = pygad.GA(
    num_generations=10,
    num_parents_mating=4,
    fitness_func=fitness_func,
    sol_per_pop=8,
    num_genes=num_genes,
    gene_space=gene_space,
    parent_selection_type="sss",
    keep_parents=1,
    crossover_type="single_point",
    mutation_type="random",
    mutation_num_genes=1
)

# Run GA
ga_instance.run()

# Show best solution
solution, solution_fitness, _ = ga_instance.best_solution()
print(f"Best solution: {solution}")
print(f"Fitness: {solution_fitness}")