GECCO Workshop on Real-Parameter Black-Box Optimization Benchmarking (BBOB 2018)
Welcome to the web page of the 8th GECCO Workshop on Real-Parameter Black-Box Optimization Benchmarking (BBOB 2018) taking place during GECCO 2018 in Kyoto, Japan.
WORKSHOP ON REAL-PARAMETER BLACK-BOX OPTIMIZATION BENCHMARKING (BBOB 2018)held as part of the2018 Genetic and Evolutionary Computation Conference (GECCO-2018)July 15–19, Kyoto, Japan
Quantifying and comparing the performance of optimization algorithms is a difficult and tedious task to achieve. Yet, it is highly crucial in order to recommend algorithms performing well in practice. The Black-box Optimization Benchmarking workshop series aims at bringing together researchers from the optimization field to discuss the latest achievements in (blackbox) optimization benchmarking as well as at gathering and sharing data of extensive numerical benchmarking results.
Open to all topics around blackbox optimization benchmarking, a substantial portion of the workshops’ past success can be attributed to the Comparing Continuous Optimization benchmarking platform (COCO, https://github.com/numbbo/coco ) that builds the basis for all BBOB workshops and that allows algorithms to be benchmarked and performance data to be visualized effortlessly. Up to now, the BBOB workshops have covered benchmarking of blackbox optimization algorithms for single- and bi-objective, unconstrained problems in exact and noisy, as well as expensive and non-expensive scenarios.
Like for the previous editions of the workshop, we provide source code in various languages (C/C++, Matlab/Octave, Java, and Python) to benchmark algorithms on three different test suites (single-objective with and without noise a well as a noiseless bi-objective suite). Postprocessing data and comparing algorithm performance is equally automatized with COCO (up to already prepared LaTeX templates for writing papers). As a new feature for the 2018 edition, we provide significantly easier access to the already benchmarked data sets such that the analysis of already available COCO data becomes simple(r).
Analyzing the vast amount of available benchmarking data (from 200+ experiments collected throughout the years) will be therefore one special focus of BBOB-2018. Given that the field of (multiobjective) Bayesian optimization received renewed interest in the recent past, we would also like to re-focus our efforts towards benchmarking algorithms for expensive problems (aka surrogate-assisted algorithms developed for limited budgets). Moreover, several classical multiobjective optimization algorithms have not yet been benchmarked on the bbob-biobj test suite, provided since 2016, such that we encourage contributions on these three following topics in particular:
analysis of existing benchmarking data
Interested participants of the workshop are invited to submit any paper around the topic of (blackbox) optimization benchmarking. These contributions might or might not use the provided LaTeX templates to visualize the performance of unconstrained single- or multiobjective black-box optimization algorithms of their choice on any of the provided testbeds. We particularly encourage submissions about algorithms from outside the evolutionary computation community as well as any papers related to topics around optimization algorithm benchmarking.
If participants wish to contribute to the BBOB workshop by submitting data sets, obtained with COCO, the tasks 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.
Note again 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 1, 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 v2.2 for running your benchmarking experiments in 2018.
Documentation of the functions used in the
is provided at http://numbbo.github.io/coco-doc/bbob-biobj/functions/ .
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 https://numbbo.github.io/coco/oldcode/bboball15.03.tar.gz 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, …
Paper submissions are expected to be done through the official GECCO submission system at https://ssl.linklings.net/conferences/gecco/ until the deadline on February 27, 2018.
In order to finalize your submission, we kindly ask you to fill in
addition the form at http://numbbo.github.io/submit where you are
supposed to provide a link to your data as well if this applies.
To upload your data to the web, you might want to use
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 for example in case you submit a paper unrelated
to the above BBOB test suites.
The data of previously compared algorithms can be found at https://numbbo.github.io/data-archive/ and is easily accessible from the python
cocoppmodule via its
Out of six submissions, the following four papers have been accepted after peer-review:
Kouhei Nishida and Youhei Akimoto: “Benchmarking the PSA-CMA-ES on the BBOB Noiseless Testbed”
Duc Manh Nguyen: “Benchmarking a Variant of the CMAES-APOP on the BBOB Noiseless Testbed”
Aurore Blelly, Matheus Felipe-Gomes, Anne Auger, and Dimo Brockhoff: “Stopping Criteria, Initialization, and Implementations of BFGS and their Effect on the BBOB Test Suite”
Aljoša Vodopija, Tea Tušar, Bogdan Filipič: “Comparing Black-Box Differential Evolution and Classic Differential Evolution”
Links to Algorithm Data
The data of all submitted experiments can be found in the list of data sets.
This year, the BBOB-2018 workshop got assigned a single session at GECCO in which the talks were scheduled according to the table below. Speakers are highlighted with a star behind the name. Please click on the provided links to download the slides.
09:30 - 09:45
The BBOBies: Introduction to Blackbox Optimization Benchmarking (slides)
09:45 - 10:05
Kouhei Nishida* and Youhei Akimoto: Benchmarking the PSA-CMA-ES on the BBOB Noiseless Testbed (slides)
10:05 - 10:25
Duc Manh Nguyen*: Benchmarking a Variant of the CMAES-APOP on the BBOB Noiseless Testbed (slides)
10:25 – 10:40
Aurore Blelly, Matheus Felipe-Gomes, Anne Auger, and Dimo Brockhoff*: Stopping Criteria, Initialization, and Implementations of BFGS and their Effect on the BBOB Test Suite (slides)
10:40 - 11:00
Aljoša Vodopija, Tea Tušar*, Bogdan Filipič: Comparing Black-Box Differential Evolution and Classic Differential Evolution
11:00 - 11:10
The BBOBies: Wrap-up and Discussion (slides)
2018-02-27 paper submission system opened
2018-03-01 release 2.2 of the COCO platform: https://github.com/numbbo/coco/releases/ (originally planned on 2018-01-05)
2018-03-27 paper and data submission deadline (not extendable!)
2018-04-10 decision notification
2018-04-24 deadline camera-ready papers
All dates are given in ISO 8601 format (yyyy-mm-dd).
Anne Auger, Inria Saclay - Ile-de-France, France
Julien Bect, CentraleSupélec, France
Dimo Brockhoff, Inria Saclay - Ile-de-France, France
Nikolaus Hansen, Inria Saclay - Ile-de-France, France
Rodolphe Le Riche, Ecole Nationale Supérieure des Mines de Saint–Etienne, France
Victor Picheny, INRA Occitanie-Toulouse, France
Tea Tušar, Jožef Stefan Institute, Ljubljana, Slovenia