RDEx-SOP: Exploitation-Biased Reconstructed Differential Evolution for Fixed-Budget Bound-Constrained Single-Objective Optimization
New evolutionary algorithm combines success-history adaptation with exploitation bias to optimize under strict evaluation limits.
A research team led by Sichen Tao has introduced RDEx-SOP, a specialized evolutionary algorithm designed for the IEEE CEC 2025 numerical optimization competition's special session (C06). The algorithm addresses bound-constrained single-objective optimization problems, which serve as key benchmarks for evaluating the robustness and efficiency of evolutionary computation methods. RDEx-SOP stands for 'Exploitation-Biased Reconstructed Differential Evolution for Fixed-Budget Bound-Constrained Single-Objective Optimization' and represents a success-history differential evolution variant with specific enhancements for competition environments.
RDEx-SOP combines three key technical components: success-history parameter adaptation that learns optimal settings during optimization, an exploitation-biased hybrid branch that focuses computational resources on promising regions, and lightweight local perturbations that help escape local optima without excessive computational cost. This combination aims to balance rapid convergence with high final solution quality under strict evaluation budgets—a critical requirement for competition settings where function evaluations are limited.
The algorithm was evaluated using the official CEC 2025 SOP benchmark suite comprising 29 diverse test functions. Performance was measured using the U-score framework, which assesses both Speed (how quickly solutions improve) and Accuracy (final solution quality) categories. Experimental results indicate that RDEx-SOP achieves strong overall performance and statistically competitive final outcomes across the benchmark set, demonstrating its effectiveness for fixed-budget optimization scenarios.
This work contributes to the ongoing development of evolutionary algorithms that must perform well under practical constraints, where computational resources are limited and solutions must be found within specific evaluation budgets. The competition-focused design makes RDEx-SOP particularly relevant for real-world optimization problems where time or computational cost represents a significant constraint.
- Combines success-history parameter adaptation with exploitation-biased hybrid branch for balanced performance
- Tested on 29 CEC 2025 benchmark functions using U-score framework (Speed and Accuracy)
- Designed specifically for fixed-budget optimization where evaluation counts are strictly limited
Why It Matters
Advances evolutionary algorithms for real-world problems where computational resources are constrained and solutions must be found quickly.