Skip to main content

A general purpose Library for Evolutionary Algorithms in Python.

Project description

LEAP: Evolutionary Algorithms in Python

Written by Dr. Jeffrey K. Bassett, Dr. Mark Coletti, and Eric Scott

Build Status Coverage Status Documentation Status

LEAP is a general purpose Evolutionary Computation package that combines readable and easy-to-use syntax for search and optimization algorithms with powerful distribution and visualization features.

LEAP's signature is its operator pipeline, which uses a simple list of functional operators to concisely express a metaheuristic algorithm's configuration as high-level code. Adding metrics, visualization, or special features (like distribution, coevolution, or island migrations) is often as simple as adding operators into the pipeline.

Using LEAP

Get the stable version of LEAP from the Python package index with

pip install leap_ec

Simple Example

The easiest way to use an evolutionary algorithm in LEAP is to use the leap_ec.simple package, which contains simple interfaces for pre-built algorithms:

from leap_ec.simple import ea_solve

def f(x):
    """A real-valued function to optimized."""
    return sum(x)**2

ea_solve(f, bounds=[(-5.12, 5.12) for _ in range(5)], maximize=True)

Genetic Algorithm Example

The next-easiest way to use LEAP is to configure a custom algorithm via one of the metaheuristic functions in the leap_ec.algorithms package. These interfaces offer you a flexible way to customize the various operators, representations, and other components that go into a modern evolutionary algorithm.

Metaheuristics are usually defined by three main objects: a Problem, a Representation, and a pipeline (list) of Operators.

Here's an example that applies a genetic algorithm variant to solve the MaxOnes optimization problem. It uses bitflip mutation, uniform crossover, and binary tournament_selection selection:

from leap_ec.algorithm import generational_ea
from leap_ec import ops, decoder, representation
from leap_ec.binary_rep import initializers
from leap_ec.binary_rep import problems
from leap_ec.binary_rep.ops import mutate_bitflip

pop_size = 5
ea = generational_ea(max_generations=10, pop_size=pop_size,

                     # Solve a MaxOnes Boolean optimization problem
                     problem=problems.MaxOnes(),

                     representation=representation.Representation(
                         # Genotype and phenotype are the same for this task
                         decoder=decoder.IdentityDecoder(),
                         # Initial genomes are random binary sequences
                         initialize=initializers.create_binary_sequence(length=10)
                     ),

                     # The operator pipeline
                     pipeline=[
                               # Select parents via tournament_selection selection
                               ops.tournament_selection,
                               ops.clone,  # Copy them (just to be safe)
                               # Basic mutation with a 1/L mutation rate
                               mutate_bitflip(expected_num_mutations=1),
                               # Crossover with a 40% chance of swapping each gene
                               ops.uniform_crossover(p_swap=0.4),
                               ops.evaluate,  # Evaluate fitness
                               # Collect offspring into a new population
                               ops.pool(size=pop_size)
                               ])

print('Generation, Best_Individual')
for i, best in ea:
    print(f"{i}, {best}")

Low-level Example

However, it may sometimes be necessary to have access to low-level details of an EA implementation, in which case the programmer can arbitrarily connect individual components of the EA workflow for maximum tailorability. For example:

from toolz import pipe

from leap_ec.individual import Individual
from leap_ec.decoder import IdentityDecoder
from leap_ec.context import context

import leap_ec.ops as ops
from leap_ec.binary_rep.problems import MaxOnes
from leap_ec.binary_rep.initializers import create_binary_sequence
from leap_ec.binary_rep.ops import mutate_bitflip
from leap_ec import util

# create initial rand population of 5 individuals
parents = Individual.create_population(5,
                                       initialize=create_binary_sequence(4),
                                       decoder=IdentityDecoder(),
                                       problem=MaxOnes())
# Evaluate initial population
parents = Individual.evaluate_population(parents)

# print initial, random population
util.print_population(parents, generation=0)

# generation_counter is an optional convenience for generation tracking
generation_counter = util.inc_generation(context=context)

while generation_counter.generation() < 6:
    offspring = pipe(parents,
                     ops.tournament_selection,
                     ops.clone,
                     mutate_bitflip(expected_num_mutations=1),
                     ops.uniform_crossover(p_swap=0.2),
                     ops.evaluate,
                     ops.pool(size=len(parents)))  # accumulate offspring

    parents = offspring

    generation_counter()  # increment to the next generation

    util.print_population(parents, context['leap']['generation'])

More Examples

A number of LEAP demo applications are found in the the example/ directory of the github repository:

git clone https://github.com/AureumChaos/LEAP.git
python LEAP/examples/advanced/island_models.py

Demo of LEAP running a 3-population island model on a real-valued optimization problem. Demo of LEAP running a 3-population island model on a real-valued optimization problem.

Documentation

The stable version of LEAP's full documentation is over at ReadTheDocs.

If you want to build a fresh set of docs for yourself, you can do so after running make setup:

make doc

This will create HTML documentation in the docs/build/html/ directory. It might take a while the first time, since building the docs involves generating some plots and executing some example algorithms.

Installing from Source

To install a source distribution of LEAP, clone the repo:

git clone https://github.com/AureumChaos/LEAP.git

And use the Makefile to install the package:

make setup

Run the Test Suite

LEAP ships with a two-part pytest harness, divided into fast and slow tests. You can run them with

make test-fast

and

make test-slow

respectively.

pytest output example

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

leap_ec-0.7.1.tar.gz (3.3 MB view hashes)

Uploaded Source

Built Distribution

leap_ec-0.7.0-py3-none-any.whl (140.1 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page