jmetalpy 1.9.0

Python version of the jMetal framework

A paper introducing jMetalPy is available at: https://doi.org/10.1016/j.swevo.2019.100598

Table of Contents

• Installation
• Usage
• Features
• Changelog
• License

Installation

You can install the latest version of jMetalPy with pip,

pip install jmetalpy  # or "jmetalpy[distributed]"

Notes on installing with pip

jMetalPy includes features for parallel and distributed computing based on pySpark and Dask.

These (extra) dependencies are not automatically installed when running pip, which only comprises the core functionality of the framework (enough for most users):

pip install jmetalpy

This is the equivalent of running:

pip install "jmetalpy[core]"

Other supported commands are listed next:

pip install "jmetalpy[dev]"  # Install requirements for development
pip install "jmetalpy[distributed]"  # Install requirements for parallel/distributed computing
pip install "jmetalpy[complete]"  # Install all requirements

Hello, world! 👋

Examples of configuring and running all the included algorithms are located in the documentation.

from jmetal.algorithm.multiobjective import NSGAII
from jmetal.operator.crossover import SBXCrossover
from jmetal.operator.mutation import PolynomialMutation
from jmetal.problem import ZDT1
from jmetal.util.termination_criterion import StoppingByEvaluations

problem = ZDT1()

algorithm = NSGAII(
    problem=problem,
    population_size=100,
    offspring_population_size=100,
    mutation=PolynomialMutation(probability=1.0 / problem.number_of_variables(), distribution_index=20),
    crossover=SBXCrossover(probability=1.0, distribution_index=20),
    termination_criterion=StoppingByEvaluations(max_evaluations=25000)
)

algorithm.run()

We can then proceed to explore the results:

from jmetal.util.solution import get_non_dominated_solutions, print_function_values_to_file, print_variables_to_file

front = get_non_dominated_solutions(algorithm.result())

# save to files
print_function_values_to_file(front, 'FUN.NSGAII.ZDT1')
print_variables_to_file(front, 'VAR.NSGAII.ZDT1')

Or visualize the Pareto front approximation produced by the algorithm:

from jmetal.lab.visualization import Plot

plot_front = Plot(title='Pareto front approximation', axis_labels=['x', 'y'])
plot_front.plot(front, label='NSGAII-ZDT1', filename='NSGAII-ZDT1', format='png')

Features

The current release of jMetalPy (v1.9.0) contains the following components:

• Algorithms: local search, genetic algorithm, evolution strategy, simulated annealing, random search, NSGA-II, NSGA-III, SMPSO, OMOPSO, MOEA/D, SMS-EMOA, MOEA/D-DRA, MOEA/D-IEpsilon, GDE3, SPEA2, HYPE, IBEA. Preference articulation-based algorithms (G-NSGA-II, G-GDE3, G-SPEA2, SMPSO/RP); Dynamic versions of NSGA-II, SMPSO, and GDE3.
• Parallel computing based on Apache Spark and Dask.
• Benchmark problems: ZDT1-6, DTLZ1-2, FDA, LZ09, LIR-CMOP, RWA, RE, unconstrained (Kursawe, Fonseca, Schaffer, Viennet2), constrained (Srinivas, Tanaka).
• Encodings: real, integer, binary, permutations.
• Operators: selection (binary tournament, ranking and crowding distance, random, nary random, best solution), crossover (single-point, SBX), mutation (bit-blip, polynomial, uniform, random).
• Quality indicators: hypervolume, additive epsilon, GD, IGD, IGD+.
• Pareto front approximation plotting in real-time, static or interactive.
• Experiment class for performing studies either alone or alongside jMetal.
• Pairwise and multiple hypothesis testing for statistical analysis, including several frequentist and Bayesian testing methods, critical distance plots and posterior diagrams.

Changelog

• [1.9.0]
  • Add RE benchmark
  • Refactor algorithm SPEA2
  • Refactor classes Solution, BinarySolution, and FloatSolution
  • Added Unimodal Normal Distribution Crossover (UNDX), BLX-Alpha, BLX-Alpha-Beta and Arithmetic crossover operators.
  • Added Levy flight and power-law mutation operators.
• [v.1.8.0]
  • Add the SMS-EMOA algorithm (based on moocore.)
  • Add the IGD+ quality indicator
  • Add new bencharmk problems:
    • eqdtlz.py
    • misc.py
    • Their reference fronts have been obtained with the Reference Set Generator method.
  • The project structure has been changed from flat to src.
  • The Hypervolume quality indicator implementation relies now on the moocore project.
• [v1.7.0] Add RWA benchmark, refactor classes BinarySolution and BinaryProblem.
• [v1.6.0] Refactor class Problem, the single-objective genetic algorithm can solve constrained problems, performance improvements in NSGA-II, generation of Latex tables summarizing the results of the Wilcoxon rank sum test, added a notebook folder with examples.
• [v1.5.7] Use of linters for catching errors and formatters to fix style, minor bug fixes.
• [v1.5.6] Removed warnings when using Python 3.8.
• [v1.5.5] Minor bug fixes.
• [v1.5.4] Refactored quality indicators to accept numpy array as input parameter.
• [v1.5.4] Added CompositeSolution class to support mixed combinatorial problems. #69

License

This project is licensed under the terms of the MIT - see the LICENSE file for details.