Skip to content
/ gaft Public

A Genetic Algorithm Framework in Python (not for production level)

License

Notifications You must be signed in to change notification settings

PytLab/gaft

Repository files navigation

GAFT

A Genetic Algorithm Framework in pyThon

Build Status Codecov Code Health platform versions Documentation Status

Introduction

GAFT is a general Python Framework for genetic algorithm computation. It provides built-in genetic operators for target optimization and plugin interfaces for users to define your own genetic operators and on-the-fly analysis for algorithm testing.

GAFT is now accelerated using MPI parallelization interfaces. You can run it on your cluster in parallel with MPI environment.

Python Support

GAFT requires Python version 3.x (Python 2.x is not supported).

Installation

  1. Via pip:

    pip install gaft
    
  2. From source:

    python setup.py install
    

If you want GAFT to run in MPI env, please install mpi4py explicitly:

pip install mpi4py

See INSTALL.md for more installation details.

Test

Run unit test:

python setup.py test

Quick start

1. Importing

from gaft import GAEngine
from gaft.components import BinaryIndividual, Population
from gaft.operators import RouletteWheelSelection, UniformCrossover, FlipBitMutation

# Analysis plugin base class.
from gaft.plugin_interfaces.analysis import OnTheFlyAnalysis

2. Define population

indv_template = BinaryIndividual(ranges=[(0, 10)], eps=0.001)
population = Population(indv_template=indv_template, size=50)
population.init()  # Initialize population with individuals.

3. Create genetic operators

# Use built-in operators here.
selection = RouletteWheelSelection()
crossover = UniformCrossover(pc=0.8, pe=0.5)
mutation = FlipBitMutation(pm=0.1)

4. Create genetic algorithm engine to run optimization

engine = GAEngine(population=population, selection=selection,
                  crossover=crossover, mutation=mutation,
                  analysis=[FitnessStore])

5. Define and register fitness function

@engine.fitness_register
def fitness(indv):
    x, = indv.solution
    return x + 10*sin(5*x) + 7*cos(4*x)

or if you want to minimize it, you can add a minimization decorator on it

@engine.fitness_register
@engine.minimize
def fitness(indv):
    x, = indv.solution
    return x + 10*sin(5*x) + 7*cos(4*x)

6. Define and register an on-the-fly analysis (optional)

@engine.analysis_register
class ConsoleOutput(OnTheFlyAnalysis):
    master_only = True
    interval = 1
    def register_step(self, g, population, engine):
        best_indv = population.best_indv(engine.fitness)
        msg = 'Generation: {}, best fitness: {:.3f}'.format(g, engine.fmax)
        engine.logger.info(msg)

7. Run

if '__main__' == __name__:
    engine.run(ng=100)

8. Evolution curve

https://github.com/PytLab/gaft/blob/master/examples/ex01/envolution_curve.png

9. Optimization animation

https://github.com/PytLab/gaft/blob/master/examples/ex01/animation.gif

See example 01 for a one-dimension search for the global maximum of function f(x) = x + 10sin(5x) + 7cos(4x)

Global maximum search for binary function

https://github.com/PytLab/gaft/blob/master/examples/ex02/surface_animation.gif

See example 02 for a two-dimension search for the global maximum of function f(x, y) = y*sin(2*pi*x) + x*cos(2*pi*y)

Plugins

You can define your own genetic operators for GAFT and run your algorithm test.

The plugin interfaces are defined in /gaft/plugin_interfaces/, you can extend the interface class and define your own analysis class or genetic operator class. The built-in operators and built-in on-the-fly analysis can be treated as an official example for plugins development.

Blogs(Chinese Simplified)

TODO

  1. ✅ Parallelization
  2. ✅ Add more built-in genetic operators with different algorithms
  3. 🏃 Add C++ backend(See GASol)

Obtain a copy

The GAFT framework is distributed under the GPLv3 license and can be obtained from the GAFT git repository or PyPI