Further Enquiries

School of Computer Science
Ingkarni Wardli Building
The University of Adelaide
SA 5005

Telephone: +61 8 8313 5586
Facsimile: +61 8 8313 4366

Competition - Evolutionary Submodular Optimisation GECCO 2022

Submodular functions play a key role in the area of optimisation as they allow to model many real-world optimisation problem. Submodular functions model a wide range of problems where the benefit of adding solution components diminishes with the addition of elements. They form an important class of optimization problems, and are extensively studied in the literature. Problems that may be formulated in terms of submodular functions include influence maximization in social networks, maximum coverage, maximum cut in graphs, sensor placement problem, and sparse regression. In recent years, the design and analysis of evolutionary algorithms for submodular optimisation problems has gained increasing attention in the evolutionary computation and artificial intelligence community.

The aim of the competition is to provide a platform for researchers working evolutionary computing methods and interested in benchmarking them on a wide class of combinatorial optimization problems. The competition will benchmark evolutionary computing techniques for submodular optimisation problems and enable performance comparison for this type of problems. It provides an idea vehicle for researchers and students to design new algorithms and/or benchmark their existing approaches on a wide class of combinatorial optimization problems captured by submodular functions.

A description of the different submodular optimization problems included in this competion can be found in Submodular-Problems.pdf

Technical details:

We will use
IOHprofiler for the competition. The problems are integrated into IOHprofiler and you can run your algorithms using it for evaluation and obtaining your results.
The problem implementations and training benchmarks will be available in IOHprofiler by 15 April 2022.

The submodular problem implementations can be found in this directory IOHprofiler Problem.
The example code for using these problem classes is located in this directory IOHprofiler Example.
Example: Example on how to use IOHprofiler.


We will evaluate all submissions on different instances of the submodular problems provided as part of the submodular problem implementations. The algorithms will be evaluated with respect to the fixed budget perspective. We will consider two categories and will determine a winner for each of the two categories.
  • In the low budget category each algorithm will be run for 10,000 fitness evaluations and the best solution obtained during the run will be used as the result.
  • In the high budget category each algorithm will be run for 100,000 fitness evaluations and the best solution obtained during the run will be used as the result.
Each algorithm will be run on each test evaluation instance 10 times in each category. Algorithms will be ranked for each considered instance. The winner is the approach that has the smallest average rank in each category, i. e. low budget or high budget. As a default, we assume that each submission should be considered for both categories. If a submission should only be considered in one of the categories, i. e. either low or high budget we ask the contributors to clearly state this in the submission email.


Submission deadline: 26 June 2022, AoE.


Aneta Neumann, University of Adelaide, Australia
Frank Neumann, University of Adelaide, Australia
Chao Qian, Nanjing University, China

Technical Support:

Viet Anh Do and IOHProfiler Development Team (Diederick Vermetten, Jacob de Nobel, Furong Ye, Hao Wang, Ofer M. Shir, Carola Doerr, Thomas Bäck)