Competition on Evolutionary Expensive Optimization for Mixed-Variable Optimization Problems (Category C)

Background

Evolutionary algorithms for expensive optimization have proven to be highly effective in addressing computationally intensive design problems across various engineering domains [1, 2]. Over the past two decades, numerous expensive evolutionary optimizers have been developed and successfully applied to a wide range of real-world scenarios [3, 4]. However, the majority of existing approaches are tailored to problems involving a single type of decision variable—either purely continuous or purely discrete [6]. With rapid advancements in engineering technologies, modern design challenges increasingly involve mixed-variable optimization problems, where continuous and categorical (or discrete) variables coexist [6, 7]. A representative example arises in automotive structural design: the growing adoption of composite materials alongside traditional metals has led to hybrid-material vehicle bodies [8]. Optimizing such structures requires simultaneous selection of material types (categorical variables) and tuning of geometric or physical parameters (continuous variables), resulting in a mixed-variable expensive optimization problem.

Competition Guidelines

Conference Information

Benchmark Problems

To foster the development of robust evolutionary optimizers for such challenging scenarios, this competition introduces a suite of 30 synthetic mixed-variable benchmark problems specifically designed for expensive optimization settings, denoted as EOPCCV-F1~EOPCCV-F30. These benchmarks are grouped into three categories, each comprising 10 decision variables:

This design comprehensively reflects the variable-type compositions commonly encountered in real-world engineering applications. The details of the benchmark problems can be found in [6], and the source code can be found at https://github.com/ZYZhang1/MiSACO-and-EOPCCV and is also released at PlatEMO (https://github.com/BIMK/PlatEMO) [10]. We also have prepared a baseline algorithm, i.e., MiSACO [6], in our source code.

Submission Requirements

Participants are encouraged to develop general-purpose algorithms capable of handling mixed-variable expensive optimization—not just specialized solvers for individual problems. Submissions may include novel algorithms, hybridizations of existing methods, or even commercial optimization software. For each algorithm, the team must provide a compressed folder (named as the algorithm's name) which contains three files:

• A Word document that records names, affiliations, and emails of participants.
• Source Code of the algorithm that implemented in PlatEMO.
• Raw data of the optimization results that produced in PlatEMO.

If you have any questions, please do not hesitate to contact Zhiyao Zhang. Submit your files directly to Zhiyao Zhang (email to zhiyao.zhang.cn@gmail.com).

Evaluation Criteria

All submissions will be evaluated fairly and uniformly by the organizers based on PlatEMO. For each of the 30 benchmark problems, the algorithm will be executed independently 20 times under a fixed computational budget (e.g., the maximum number of expensive function evaluations is set to 600). The mean of the final objective values obtained across the 20 runs will be used as the primary performance metric. This approach accounts for the stochastic nature of most optimization algorithms and ensures a more reliable and statistically sound comparison. The overall ranking will consider average performance across all three problem categories to assess both robustness and generalization capability.

We invite researchers and practitioners to contribute innovative solutions to this emerging and practically significant challenge in computational intelligence.

Awards

Organizing Team

References

1. Jin, Y. (2011). Surrogate-assisted evolutionary computation: Recent advances and future challenges. Swarm and Evolutionary Computation, 1(2), 61-70.
2. Ong, Y. S., Nair, P. B., & Keane, A. J. (2003). Evolutionary optimization of computationally expensive problems via surrogate modeling. AIAA journal, 41(4), 687-696.
3. Zhang, Z., Wang, Y., Sun, G., Pang, T., & Tang, K. (2025). Constrained probabilistic Pareto dominance for expensive constrained multiobjective optimization problems. IEEE Transactions on Evolutionary Computation, 29(4), 1138-1152.
4. Wang, Y., Yin, D. Q., Yang, S., & Sun, G. (2019). Global and local surrogate-assisted differential evolution for expensive constrained optimization problems with inequality constraints. IEEE transactions on cybernetics, 49(5), 1642-1656.
5. Liu, J., Wang, Y., Sun, G., & Pang, T. (2022). Multisurrogate-assisted ant colony optimization for expensive optimization problems with continuous and categorical variables. IEEE Transactions on Cybernetics, 52(11), 11348-11361.
6. Liu, Y., & Wang, H. (2023). Surrogate-assisted hybrid evolutionary algorithm with local estimation of distribution for expensive mixed-variable optimization problems. Applied Soft Computing, 133, 109957.
7. Wang, Z., Xie, L., Li, G., Gao, W., Gong, M., & Wang, L. (2025). Customized evolutionary expensive optimization: Efficient search and surrogate strategies for continuous and categorical variables. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 55(3), 2196-2210.
8. Tan, S., Wang, Y., Sun, G., & Pang, T. (2025). A Surrogate-Assisted High-Dimensional Mixed-Variable Evolutionary Framework and Its Application to Vehicle Lightweighting Design. IEEE Transactions on Evolutionary Computation.
9. Tian, Y., Cheng, R., Zhang, X., & Jin, Y. (2017). PlatEMO: A MATLAB platform for evolutionary multi-objective optimization. IEEE Computational Intelligence Magazine, 12(4), 73-87.