A Longitudinal Analysis of the CEC Single-Objective Competitions (2010-2024) and Implications for Variational Quantum Optimization
A 2026 analysis shows how a 2014 benchmark change forced AI to evolve for quantum computing.
A team of researchers led by Vojtěch Novák has published a comprehensive historical analysis of the IEEE Congress on Evolutionary Computation (CEC) single-objective optimization competitions from 2010 to 2024. The paper, posted on arXiv, identifies a pivotal moment in 2014 when competition organizers introduced benchmark functions with dense rotation matrices. This technical change created parameter non-separability, effectively filtering out coordinate-dependent optimization algorithms like Particle Swarm Optimization (PSO) and Genetic Algorithms (GA). The shift established long-term dominance for Differential Evolution variants, particularly the L-SHADE algorithm, which could preserve rotational invariance in its search vectors.
Post-2020, the analysis reveals the winning strategy evolved further into building high-complexity hybrid optimizers. These modern solvers combine multiple mechanisms—such as Eigenvector Crossover, Societal Sharing, and Reinforcement Learning—to maximize ranking stability across a diverse and challenging benchmark suite. The researchers' key insight is that the structural complexity of these modern CEC benchmarks closely mirrors the optimization landscapes faced by Variational Quantum Algorithms (VQAs). Therefore, the adaptive capabilities honed by AI over 15 years of intense competition are directly applicable to a critical quantum computing task: tuning the parameters of quantum circuits for practical applications like chemistry simulation or optimization.
- A 2014 benchmark change with dense rotation matrices made PSO and GA obsolete, cementing Differential Evolution (L-SHADE) as the dominant algorithm.
- Post-2020 winners are complex hybrid systems combining techniques like Eigenvector Crossover and Reinforcement Learning for stability.
- The study concludes the evolved AI solvers possess the specific adaptive capabilities needed to optimize Variational Quantum Algorithms (VQAs).
Why It Matters
It provides a roadmap for using battle-tested classical AI to solve the critical parameter-tuning problem holding back practical quantum algorithms.