Open-source optimization for Python

Let evolution
find a better solution.

Build genetic algorithms, search complex solution spaces, and optimize machine-learning models with a library that stays out of your way.

Get started ↗ View on GitHub

pip install pygad

Python 3.8+ · BSD-3-Clause · Lightweight core

SEARCH SPACEILLUSTRATION

GENERATION80

BEST SOLUTIONImproving ↗

An illustrative view of candidates evolving toward better fitness.

⌘ Python-first API↗ Single & multi-objective◉ Open source◎ Keras & PyTorch tools

Built for real search problems

Flexible by design.
Simple to get moving.

Start with a fitness function. Shape the search with the operators and constraints your problem needs.

01 / OPTIMIZE

One goal or many

Optimize one objective or return multiple fitness values for multi-objective search with NSGA-II and NSGA-III.

Explore multi-objective optimization ↗

02 / CUSTOMIZE

Make the search yours

Choose parent selection, crossover and mutation methods; add custom operators, constraints and gene spaces.

See the available operators ↗

03 / INTEGRATE

Optimize ML models

Use Keras and PyTorch integrations to represent model parameters as chromosomes and search for better weights.

Explore Keras integration ↗

04 / SCALE

Spend compute wisely

Batch fitness evaluations, reuse fitness values, or parallelize intensive calculations with threads or processes.

Read about performance ↗

Your first experiment

Define what
“better” means.

Write a fitness function for your problem, set the search size, and let PyGAD evolve candidate solutions.

Installpip install pygad

Define fitnessScore a candidate solution

Run and inspectExplore the best solution found

Follow the getting-started guide →

# optimize.py
import pygad

def fitness_func(ga_instance, solution, solution_idx):
    return 1.0 / (abs(sum(solution) - 42) + 1e-6)

ga = pygad.GA(
    num_generations=100,
    num_parents_mating=4,
    fitness_func=fitness_func,
    sol_per_pop=10,
    num_genes=6,
)
ga.run()

solution, fitness, _ = ga.best_solution()
print(solution, fitness)