• Amir H. Gandomi

  Faculty of Engineering & Information Technology, University of Technology Sydney, Ultimo, Australia, and University Research and Innovation Center (EKIK), Obuda University, Budapest, Hungary.
  Email: Gandomi@uts.edu.au

• Kalyanmoy Deb

  Department of Electrical and Computer Engineering, Michigan State University, East Lansing, USA.
  Email: kdeb@egr.msu.edu

• Danial Yazdani

  Data Science Institute, University of Technology Sydney, Ultimo, Australia.
  Emails: danial.yazdani@uts.edu.au and danial.yazdani@gmail.com

• Rohit Salgotra

  Faculty of Physics & Applied Computer Science, AGH University of Science & Technology, Poland.
  Email: rohit.salgotra@agh.edu.pl

• Mohammad Nabi Omidvar

  School of Computing, University of Leeds, and Leeds University Business School, Leeds LS2 9JT, UK.
  Email: m.n.omidvar@leeds.ac.uk

This competition invites researchers to test the mettle of their global optimization algorithms against a meticulously curated set of 24 problem instances from the Generalized Numerical Benchmark Generator (GNBG) [1]. This test suite spans a wide array of problem terrains, from smooth unimodal landscapes to intricately rugged multimodal realms. The suite encompasses:

• Unimodal instances (f₁ to f₆),
• Single-component multimodal instances (f₇ to f₁₅), and
• Multi-component multimodal instances (f₁₆ to f₂₄).

With challenges that include various degrees of modality, ruggedness, asymmetry, conditioning, variable interactions, basin linearity, and deceptiveness, the competition provides a robust assessment of algorithmic capabilities. But this competition is not just about finding optimal solutions. It is about understanding the journey to these solutions. Participants will decipher how algorithms navigate deceptive terrains, traverse valleys, and adapt to the unique challenges posed by each instance. In essence, it is a quest for deeper insights into optimization within complex numerical landscapes. We warmly invite researchers to partake in this competition and subject their global optimization algorithms to this rigorous test.

• The competition is open to all researchers and practitioners in the field of continuous numerical optimization.
• Competitors can participate with either newly proposed algorithms (unpublished) or their previously published algorithms.
• Winners and runners-up will receive certificates from the conference.
• It is not required to attend the conference or register in order to participate in the competition.
• No modifications are permitted to parameter settings of the instances in the '.mat' files.
• Algorithms' parameter values must be consistent across all problem instances. Parameter tuning tailored for individual instances is prohibited.
• Problem instances must be treated as blackboxes. Direct use of GNBG's internal parameters in algorithms is forbidden.
• Winners will be required to share their algorithm's source code for result verification. This code will remain confidential and won't be published.

To assess the performance of optimization algorithms in this competition, three performance indicators are used. These indicators will be calculated based on the outcomes of 31 independent runs for each algorithm on every problem instance:

1. Average Absolute Error :
   This metric is calculated as the mean of the absolute errors of the best-found solutions across the 31 runs. It reflects the algorithm's accuracy and consistency in finding solutions close to the global optimum
2. Average Function Evaluations (FEs) to Acceptance Threshold :
   This criterion measures the mean number of FEs required for the algorithm to find a solution with an absolute error smaller than 10^-8. It provides insight into the efficiency and convergence speed of the algorithm.
3. Success Rate :
   Defined as the percentage of runs in which the algorithm successfully finds a solution with an absolute error less than 10^-8. This rate is indicative of the algorithm’s reliability and robustness in consistently reaching high-accuracy solutions.