Get from Ico-github-logo

Highlights from
Test functions for global optimization algorithms

  • ackley.m
  • beale.m
  • bird.m
  • booth.m
  • bukin4.m
  • bukin6.m
  • carromtable.m
  • chichinadze.m
  • crossfunc.m
  • crossintray.m
  • crosslegtable.m
  • crownedcross.m
  • cube.m
  • easom.m
  • eggholder.m
  • ezimage(fun)
    Please report bugs and inquiries to:
  • giunta.m
  • goldsteinprice.m
  • griewank.m
  • helicalvalley.m
  • himmelblau.m
  • holdertable.m
  • leon.m
  • levi13.m
  • matyas.m
  • mccormick.m
  • modschaffer1.m
  • modschaffer2.m
  • modschaffer3.m
  • modschaffer4.m
  • penholder.m
  • powell.m
  • rastrigin.m
  • rosenbrock.m
  • schweffel.m
  • sinenvsin.m
  • sixhumpcamel.m
  • styblinskitang.m
  • testtubeholder.m
  • threehumpcamel.m
  • trigonometric.m
  • wood.m
  • zettl.m
  • View all files

5.0

5.0 | 1 rating Rate this file 62 Downloads (last 30 days) File Size: 44.8 KB File ID: #23147
image thumbnail

Test functions for global optimization algorithms

by

 

28 Feb 2009 (Updated )

Test functions for global optimization algorithms

| Watch this File

File Information
Description

This is a set of test functions which can be used to test the effectiveness of global optimization algorithms. Some are rather easy to optimize (rosenbruck, leon, ...), others next to impossible (crosslegtable, bukin6, ...).
All the test-functions are taken from either [1], [2] or [3] (see below). All functions may be called in two ways:
[dims, lb, ub, sol, fval_sol] = fun
(e.g., no input arguments) This returns the number of dimensions of the function, the default lower and upper bounds, the solution vectors for all global minima and the corresponding function values. To calculate the function value for input X, use:

val = fun( [x1, x2, ..., xn] )

with the dimension [n] depending on the specific function [fun] (for most functions, n=2). Note the single vector argument--this is done in order to easily insert the function into a global optimizer that inserts a [N x n] matrix of trial vectors in these functions.

I also included a function to display most of the functions. This is called EZIMAGE, and can be called with a function handle argument:

ezimage(@himmelblau) (to plot the himmelblau function)
ezimage(@sinenvsin) (see screenshot)
...

or just as-is:

ezimage

which lists all functions and waits for user input. This is meant to get a first impression of what the challenges are the test function has to offer.

FUTURE WORK:
- constrained single-objective functions
- (constrained ) multi-objective functions

sources:
[1] Mishra, Sudhanshu. "Some new test functions for global optimization and performance of repulsive particle swarm method". MPRA, 23rd august 2006. http://mpra.ub.uni-muenchen.de/2718/

[2] Z.K. Silagadze. "Finding two-dimensional peaks". 11th mar 2004. arXiv preprint: arXiv:physics/0402085v3

[3] W. Sun, Ya-X. Yuan. "Optimization theory and Methods. Nonlinear Programming". Springer verlag, 2006. ISBN-13:978-0-387-24975-9.

If you find this work useful, please consider a donation:
https://www.paypal.com/cgi-bin/webscr?cmd=_s-xclick&hosted_button_id=6G3S5UYM7HJ3N

Acknowledgements

This file inspired Evolution Strategies (Es).

MATLAB release MATLAB 7.9 (R2009b)
Other requirements None that I know of
Tags for This File   Please login to tag files.
Please login to add a comment or rating.
Comments and Ratings (1)
22 Mar 2012 Gilberto Alejandro Ortiz Garcia  
Updates
11 Feb 2010

- updated all functions to automate finding its dimensions/bounds
- cleaned up EZIMAGE() , and made it suitable for future extentions

14 Feb 2014

- Corrected bug in leon function (square -> cube)
- Contact info updated

17 Jun 2014

[linked to Github]

20 Jun 2014

Fixed all bugs found by Jeffrey Larson (thanks!)

Contact us