pymop 0.2.1

Optimization Test Problems

pymop - Optimization Test Problems

Installation

The test problems are uploaded to the PyPi Repository.

pip install pymop

For the current development version:

git clone https://github.com/msu-coinlab/pymop
cd pymop
python setup.py install

Problems

In this package single- as well as multi-objective test problems are included:

• Single-Objective:

  • Ackley

  • BNH

  • Griewank

  • Knapsack

  • Schwefel

  • Sphere

  • Zakharov

• Multi-Objective:

  • ZDT 1-6

  • DTLZ 1-7

  • WFG 1-9

  • Carside Impact

  • BNH

  • Kursawe

  • OSY`

  • TNK

  • Welded Beam

Usage

def evaluate():
    import numpy as np

    # initialize it with the necessary parameters
    from pymop.problems.dtlz import DTLZ1
    problem = DTLZ1(n_var=10, n_obj=3)

    # evaluation function returns by default two numpy arrays - objective function values and constraints -
    # as input either provide a vector
    F, G = problem.evaluate(np.random.random(10))

    # or a whole matrix to evaluate several solutions at once
    F, G = problem.evaluate(np.random.random((100, 10)))

    # if no constraints should be returned
    F = problem.evaluate(np.random.random((100, 10)), return_constraint_violation=False)

    from pymop.problems.welded_beam import WeldedBeam
    F, CV = WeldedBeam().evaluate(np.random.random((100, 4)), return_constraint_violation=True)


def plot():
    from pymop import plot_problem_surface, Ackley
    plot_problem_surface(Ackley(n_var=1), 200)
    plot_problem_surface(Ackley(n_var=2), 200, plot_type="wireframe")
    plot_problem_surface(Ackley(n_var=2), 200, plot_type="contour")

Implementation

All problems are implemented to efficiently evaluate multiple input points at a time. Therefore, the input can be a n x m dimensional matrix, where n is the number of points to evaluate and m the number of variables.

Contributors

Julian Blank

Yash Prasad

Contact

Feel free to contact me if you have any question:

Julian Blank (blankjul [at] egr.msu.edu)

Michigan State University

Computational Optimization and Innovation Laboratory (COIN)

East Lansing, MI 48824, USA

Changelog

0.2.1

• First official release providing a bunch of test problems

• Some redesign of classes compared to early versions

• Added trust_2d problems