GNBG-II Competition GECCO 2025 & CEC 2025
🏆 Competition • 24 GNBG-II instances • 30 runs each • Black-box

GNBG-II Numerical Global Optimization

A property-controlled test suite spanning unimodal to multi-component multimodal landscapes (conditioning, asymmetry, interactions, basin response, deceptiveness).

Explore feed How to submit
Stop criteria: f1–f15: 500k FEs • f16–f24: 1M FEs Code: MATLAB / Python / C++

News

Calls, deadlines, and quick links.

News

Benchmark code (MATLAB / Python / C++)

GNBG-II instances + wrappers. Treat instances as black-box.

Open GitHub What’s inside?
Note: keep parameters unchanged across all GNBG-II instances.
Call

Call for abstracts

Submit your algorithm entry and short abstract describing the method, settings, and reproducibility details.

  • Include algorithm name + short description.
  • Provide parameter settings (fixed across all instances).
  • Attach result files + runnable code package for verification.
Submission details Contact organizers
Dates

Important dates

  • Submission deadline: 15 January 2025
  • Verification period: late January 2025
  • Results announcement: February 2025
Separate submissions are required for the CEC and GECCO. All abstracts must be submitted no later than 31 March 2025, and full papers must be submitted by the conference proceedings deadline.

Overview

Overview

Competition overview

24 GNBG-II generated instances: unimodal → single-component multimodal → multi-component multimodal.

This competition challenges researchers to evaluate the performance of their global optimization algorithms on a carefully crafted set of 24 problem instances generated using the Generalized Numerical Benchmark Generator (GNBG). The test suite encompasses a diverse range of optimization landscapes, spanning from smooth unimodal surfaces to highly intricate and rugged multimodal terrains. The newly designed test suite follows the same baseline function as used in GECCO 2024 Competitions, but the problems instances have been changed to add further complexity in the basic problems. This test suite spans a wide array of problem terrains, from smooth unimodal landscapes to intricately rugged multimodal realms.

Reference: Rohit Salgotra, Amir H. Gandomi, Kalyanmoy Deb (2025). Numerical Global Optimization Competition on GNBG-II generated Test Suite. In Proceedings of the Genetic and Evolutionary Computation Conference Companion.

Submission Details

Rules

Rules & compliance

  • Open to all researchers in continuous numerical optimization.
  • Unpublished or previously published algorithms are allowed.
  • No per-instance tuning; parameters must remain consistent across all instances.
  • No use of GNBG internal parameters (instances are black-box).
  • No changes to instance .mat parameter settings.
  • Winners must share source code for verification (kept confidential).
Evaluation

Evaluation metrics

Average absolute error:
Mean error of best solutions across 30 runs.

Average FEs to threshold:
Mean FEs to reach absolute error < 1e-8.

Success rate:
Percentage of runs reaching absolute error < 1e-8.

Submission

Submission package

One zipped folder named with your algorithm.

Include

  • Documentation: title, authors, affiliations, emails, summary + mean±std tables.
  • 24 result files: f10.dat etc.
  • Each file contains 30 runs with 2 columns: absolute error and FEs-to-threshold.

Submit to

Deadline: 15 January 2025 (update if needed)

Results

Results

Competition results (CEC 2025)

Final ranking based on total weighted rank across all test cases.

Rank
Algorithm
Score
1
AG-GEAs
48.12
2
Modular CMAES
41.62
3
E-SHADE
30.14
Download run data Winner code

Organizing Committee

Rohit Salgotra
Rohit Salgotra
Research Assitant Professor
AGH University of Krakow • Poland
Kalyanmoy Deb
Kalyanmoy Deb
Professor
Michigan State University • USA
Amir H Gandomi
Amir H Gandomi
Professor
University of Technology Sydney • Australia

Contact

Email
dsmlossf@gmail.com
Follow Us