GECCO Workshop on Real-Parameter Black-Box Optimization Benchmarking (BBOB 2017)¶
Welcome to the web page of the 7th GECCO Workshop on Real-Parameter Black-Box Optimization Benchmarking (BBOB 2017) with a continued focus on bi-objective problems and which will take place during GECCO 2017.
WORKSHOP ON REAL-PARAMETER BLACK-BOX OPTIMIZATION BENCHMARKING (BBOB 2017)hold as part of the2017 Genetic and Evolutionary Computation Conference (GECCO-2017)July 15–19, Berlin, Germany
|register for news||Coco quick start (scroll down a bit)||latest Coco release|
Quantifying and comparing the performance of optimization algorithms is a difficult and tedious task to achieve. The Coco platform provides tools to ease this process for single-objective noiseless and noisy problems and for bi-objective noiseless problems by: (1) an implemented, well-motivated benchmark function testbed, (2) a simple and sound experimental set-up, (3) the generation of output data and (4) the post-processing and presentation of the results in graphs and tables.
In 2017, we provide a new extension of the 2016 bi-objective test suite to 92 problems overall—addressing certain issues raised at last year’s BBOB workshop—and continue to support all previously known ones. Overall, we now provide the following test suites:
bbobcontaining 24 noiseless functions,
bbob-noisycontaining 30 noisy functions [^1],
bbob-biobjcontaining 55 noiseless, bi-objective functions, and
bbob-biobj-extcontaining 92 noiseless, bi-objective functions
The tasks for participants are as usual: run your favorite single- or multiobjective black-box optimizer (old or new) by using the wrappers provided (in C/C++, Python, Java, and Matlab/Octave) and run the post-processing procedure (provided as well) that will generate automatically all the material for a workshop paper (ACM compliant LaTeX templates available). A description of the algorithm and the discussion of the results completes the paper writing.
We encourage particularly submissions related to the new
and algorithms for handling bi-objective problems including also algorithms
from outside the evolutionary computation community. Submissions related to
the previously available
are more than welcome.
Please note that any other submission, related to black-box optimization benchmarking of continuous optimizers will be welcome as well. The submission section below gives a few examples of subjects of interest.
During the workshop, algorithms, results, and discussions will be presented by the participants. An overall analysis and comparison of all submitted algorithm data is going to be accomplished by the organizers and the overall process will be critically reviewed.
Updates and News¶
Get updated about the latest news regarding the workshop and releases and bugfixes of the supporting NumBBO/Coco platform, by registering at http://numbbo.github.io/register.
Basis of the workshop is the Comparing Continuous Optimizer platform (https://github.com/numbbo/coco), now rewritten fully in ANSI C with other languages calling the C code. Languages currently available are C, Java, MATLAB/Octave, and Python.
Most likely, you want to read the Coco quick start (scroll down a bit). This page also provides the code for the benchmark functions, for running the experiments in C, Java, Matlab, Octave, and Python, and for postprocessing the experiment data into plots, tables, html pages, and publisher-conform PDFs via provided LaTeX templates. Please refer to http://numbbo.github.io/coco-doc/experimental-setup/ for more details on the general experimental set-up for black-box optimization benchmarking.
The latest (hopefully) stable release of the Coco software can be downloaded as a whole here. Please use at least version v2.0 for running your benchmarking experiments in 2017.
Documentation of the functions used in the
for BBOB 2017 are provided at http://numbbo.github.io/coco-doc/bbob-biobj/functions/ .
[^1] Note that the current release of the new Coco platform does not contain the original noisy BBOB testbed yet, such that you must use the old code at http://coco.gforge.inria.fr/doku.php?id=downloads for the time being if you want to compare your algorithm on the noisy testbed.
We encourage any submission that is concerned with black-box optimization benchmarking of continuous optimizers, for example papers that:
- describe and benchmark new or not-so-new algorithms on one of the above testbeds,
- compare new or existing algorithms from the COCO/BBOB database[^2],
- analyze the data obtained in previous editions of BBOB[^2], or
- discuss, compare, and improve upon any benchmarking methodology for continuous optimizers such as design of experiments, performance measures, presentation methods, benchmarking frameworks, test functions, ...
Submissions are expected to be done through the submission form at: http://numbbo.github.io/submit.
To upload your data, you might want to use https://zenodo.org/ which
offers uploads of data sets up to 50GB in size or any other provider
of online data storage.
Please let us know briefly in the mandatory
Data field, why you do
not provide any data in case you submit a paper unrelated to the above
BBOB test suites.
[^2] The data of previously compared algorithms can be found at
http://coco.gforge.inria.fr/doku.php?id=algorithms-biobj for the
bbob-biobj test suite and at http://coco.gforge.inria.fr/doku.php?id=algorithms
bbob-noisy test suites.
- 01/28/2017 release 2.0 of the Coco platform for first tests: https://github.com/numbbo/coco/releases/
- 03/07/2017 expected release of the Coco software with the final functionality to run experiments
- 03/31/2017 paper and data submission deadline
- 04/17/2017 decision notification
- 04/24/2017 deadline camera-ready papers
- 07/15/2017 workshop
- Anne Auger, Inria Saclay - Ile-de-France, France
- Dimo Brockhoff, Inria Saclay - Ile-de-France, France
- Nikolaus Hansen, Inria Saclay - Ile-de-France, France
- Dejan Tušar, Inria Saclay - Ile-de-France, France
- Tea Tušar, Jozef Stefan Institute, Ljublana, Slovenia