Experimental comparison of six population-based algorithms for continuous black box optimization.

Area of Science:

• Optimization algorithms
• Computational intelligence
• Numerical analysis

Background:

• Direct search methods like Nelder-Mead simplex search are widely used.
• Evolutionary computation offers advanced population-based optimization techniques.
• Benchmarking is crucial for evaluating and comparing optimization algorithm performance.

Purpose of the Study:

• To comprehensively compare six population-based methods for real-valued black box optimization.
• To evaluate the performance of established and recent optimization algorithms.
• To identify the most effective and efficient algorithm for real-valued black box optimization problems.

Main Methods:

• Utilized the Comparing Continuous Optimizers (COCO) methodology for a standardized comparison.
• Evaluated Nelder-Mead simplex search, POEMS, G3PCX, Cauchy EDA, BIPOP-CMA-ES, and CMA-ES.
• Assessed algorithms based on success rates and computational speed.

Main Results:

• BIPOP-CMA-ES achieved the highest success rates across various optimization problems.
• BIPOP-CMA-ES was frequently among the fastest algorithms evaluated.
• Other algorithms showed mixed performance, with Cauchy EDA and POEMS often being slow.

Conclusions:

• BIPOP-CMA-ES is a highly effective and efficient algorithm for real-valued black box optimization.
• The COCO methodology provides a robust framework for comparing optimization algorithms.
• Further research may explore hybrid approaches or parameter tuning for other algorithms.