pyparadiseo 0.5

Python version of the ParadisEO framework for metaheuristic optimization

For the documentation of PyParadiseo see this.

This README only gives a very short introduction.

Table of Contents

• Installation
• Usage
• Components

Installation

The easiest way to get pyparadiseo is to install it via pip. Currently the following Python versions are supported: 3.6, 3.7, 3.8, 3.9, 3.10

1) Recommended : install with pip

You can install pyParadiseo with pip

    pip install pyparadiseo

2) Build from source

To build pyParadiseo, you'll need to have a few prerequisites installed on your system and set the corresponding paths in setup.py and CMakeLists.txt

To compile the binary extension you need:

• cmake >= 3.14
• python3 >= 3.6
• boost-python
• boost-numpy

If you want to build pyparadiseo from source, the easiest should be to use this manylinux2014_x86_64 Docker image with installed prerequisites.

Getting Started

The documentation of PyParadiseo is available here

Example of running EO's simple GA (SGA) for the One-Max test problem

from pyparadiseo import config

# problem dependent
from pyparadiseo import population
from pyparadiseo import initializer
from pyparadiseo import evaluator
from pyparadiseo import operator

from pyparadiseo.eo import algo,select_one,continuator

DIM=20
POP_SIZE=25
MAX_GEN=100

if __name__ == "__main__":
    #set global solution type 'bin'
    config.set_solution_type('bin')

    #make pyparadiseo evaluator from python function
    eval = evaluator.fitness(lambda sol: sum(sol))

    #generate and evaluate population
    pop=population.from_init(POP_SIZE,initializer.random(DIM))
    evaluator.pop_eval_from_fitness(eval)(pop,pop)

    #assemble simple GA
    sga = algo.simpleGA(
        select_one.det_tournament(4),
        operator.OnePtBitCrossover(),.1,
        operator.DetBitFlip(),.7,
        eval,
        continuator.max_generations(MAX_GEN)
    )
    # #run algo on pop and print best individual
    sga(pop)
    print(pop.best())