IEEE CEC 2023 will host competitions to stimulate research in evolutionary computation, promote fair evaluations, and attract congress participants. The following table includes accepted competitions. Abstracts are presented below. Please visit the referenced website or contact the competition hosts via the given e-mail address to receive specific information on participating in the competition and submitting your results.
Dynamic Constrained Multiobjective Optimization
In the past decade, dynamic constrained multiobjective optimization has attracted increasing research interest. The problem is widely-spread in real-world applications, such as scheduling optimization, and resource allocation, which involve time-varying multiobjective and constraints. More especially, the corresponding dynamic constrained multiobjective optimization problems (DCMOPs) contain more complex characteristics and special difficulties than those of dynamic multiobjective optimization problems or constrained multiobjective optimization problems. To promote the research on dynamic constrained multiobjective optimization (DCMO), 10 benchmark functions are developed, covering diverse characteristics which exactly represent different real-world scenarios, for example, continuity-disconnection, time-dependent PF/PS geometries, dynamic infeasible region, small feasible region, and so on. Based on the test suite with various characteristics, researchers can better understand the strengths and weaknesses of DCMOEAs, stimulating the research on dynamic constrained multiobjective optimization.
Evolutionary Multi-task Optimization
Evolutionary multitasking opens up new horizons for researchers in the field of evolutionary computation. It provides a promising means to deal with the ever-increasing number, variety, and complexity of optimization tasks. More importantly, rapid advances in cloud computing could eventually turn optimization into an on-demand service hosted on the cloud. In such a case, a variety of optimization tasks would be simultaneously executed by the service engine where evolutionary multitasking may harness the underlying synergy between multiple tasks to provide consumers with faster and better solutions. The test suites for multi-task single-objective optimization (MTSOO) and multi-task multi-objective optimization (MTMOO) each contain nine MTO complex problems and ten 50-task MTO benchmark problems. Potential participants in this competition may target either or both of MTSOO and MTMOO while using all benchmark problems in the corresponding test suites.
Single Objective Bound Constrained Numerical Optimization
Single-objective numerical optimization is an important class of problems to be solved. All new evolutionary and swarm algorithms are tested on single-objective benchmark problems. The aim of this competition is to test the algorithms fairly, and automatically online. The competitors submit their algorithms through the online competition system.
Constrained Multimodal Multiobjective Optimization
The aim of this competition is to promote research on constrained multimodal multiobjective optimization (CMMO) and hence motivate researchers to formulate real-world practical problems. In this competition, multiple sets of CMMO test problems of different properties are presented, such as problems with different shapes of CPSs and CPFs, different relationships between UPSs and CPSs, scalable number of CPSs, decision variables, and objectives. In addition, a fair and appropriate evaluation criterion and reference data are given to assess the performance of different CMMO algorithms. We encourage all researchers to test their algorithms on the CEC’23 test suite and to submit their papers to a special session on Constrained Multimodal Multiobjective Optimization.
Large-scale Continuous Optimization for Non-contact Measurement
In this competition, there are two tracks: large-scale continuous single- and multi-objective optimization in two non-contact measurement cases. We carefully select six LSOPs for each track from two tasks, i.e., non-contact voltage measurement for multiconductor systems (NVM) and non-contact current measurement for multiconductor systems (NCM). Participants are encouraged to develop the algorithm to solve this type of optimization problem, not just a specific one of them. Participants may propose a new optimization algorithm or utilize a hybrid form of previously proposed algorithms. However, it must be restricted to the field of evolutionary computing.
Evolutionary Computation in the Energy Domain: Operation and Planning Applications
This CEC 2023 competition proposes two tracks in the energy domain: (1) Risk-based optimization of aggregators’ day-ahead energy resource management (ERM) considering the uncertainty of high penetration of distributed energy resources (DER). This testbed represents a centralized day-ahead ERM in a smart grid with a 13-bus distribution network using a 150-scenario case study with 10 scenarios considering extreme events (high impact, and low probability). (2) Transmission Network Expansion Planning. Long-term transmission network expansion planning (TNEP) is a classic problem of power systems. The objective is to find the optimal expansion plan that identifies the transmission lines that must be installed in the system to allow a proper operation within a predefined planning horizon with the lowest investment cost. The optimal expansion plan should define where and how many lines should be installed.
Competition on Multiobjective Neural Architecture Search
The advancement of neural architecture search (NAS) has facilitated the automation of deep learning network design, leading to improved performance on various challenging computer vision tasks. Due to the black-box nature of NAS, the complex properties from the optimization point of view (e.g., discrete decision variables, multimodal and noisy fitness landscapes, expensive and many objectives, etc.) pose a great challenge to EMO algorithms. In the Multiobjective Neural Architecture Search Competition, participants should use an EMO algorithm designed by themselves to solve the given NAS tasks. The algorithms used can be brand-new or adapted from previously proposed algorithms but must be relevant to the field of evolutionary computation. We provide an end-to-end pipeline, dubbed EvoXBench, to generate benchmark test suites for EMO algorithms to run efficiently without requiring GPUs or Pytorch/Tensorflow.
Competition on Seeking Multiple Optima in Dynamic Environments
Dynamic multimodal optimization problems (DMMOPs) are a class of dynamic problems with multiple optima (sometimes including the accepted local optima) in each environment. Many real-world optimization problems have both dynamic and multimodal properties, such as dynamic economic dispatch problems, dynamic load balancing problems, and so on. The objectives and/or constraints of these problems vary over time, requiring the optimization algorithms to fast-track as many optimal solutions as possible in each environment. This allows decision-makers to pick out the most satisfactory optimal solution in each environment according to their experiences and preferences, or to turn to a different solution if the current one cannot work well. In this competition, benchmarks are constructed by 8 multimodal functions and 8 change modes.
A Sandbox for Teaching and Learning in CI for Pre-University and Undergraduate Students
The idea is to create a young student workshop with an associated competition during CEC 2023 inspired by the F.I.R.S.T. robotics competition. Before or during the first hours of the event, students will receive learning materials and guidance from tutors to learn about a CI-related topic. Later students will be enabled to test learned materials in a simple real-world application. Namely, programming robots to solve various tasks.