This is an Octave/Matlab* (see note below) toolbox that implements the DTLZ multi-objective scalable test problems [1]. If you work with Multi-objective Evolutionary Algorithms (such as NGSA-II, SPEA-2, DEMO etc.) and wishes to test your algorithm with an arbitrary number of objectives, chances are that this toolbox will be useful for you.
Currently this repository implements:
- DTLZ functions from DTLZ1 to DTLZ7 with any number of objectives M ≥ 2;
- A code to measure the distance from a given solution x to the Pareto-optimal set of a DTLZ function (this approach is better explained in [2], chapter 6);
- A code to compute the ideal and Nadir solutions of these functions (if the
Nadir computation of the DTLZ7 seems too arbitrary, check the small documentation
in the
docsfolder explaining its derivation).
- Octave ≥ 4.0;
To use these functions, download the folder Octave-Matlab and put it into your
path. Each m file has a helper showing how to use the code, so, in doubt,
type help filename in the console and check the examples or simply read the
code.
The code was developed and tested in Octave. In my opinion, besides being free, it is now more than capable of replacing Matlab, which is just too expensive for use outside a large company (and even in this case I can argue that Octave is also enough). Even so, I tried to make the code compatible with both softwares. However, since I have no Matlab copy to check, if it does break, I would be honored to receive a bug or a (even better) pull request helping to solve it.
- Kalyanmoy Deb, Lothar Thiele, Marco Laumanns, and Eckart Zitzler. Scalable test problems for evolutionary multiobjective optimization. Technical Report 1990, 2005.
- Fillipe Goulart. Preference-guided Evolutionary Algorithms for Optimization with Many Objectives. Master’s thesis, Universidade Federal de Minas Gerais, 2014.