Hot Off the Press: Benchmarking Derivative-Free Global Optimization Algorithms under Limited Dimensions and Large Evaluation Budgets

conference paper

en

This Hot Off the Press paper provides a brief summary of our recent work "Benchmarking Derivative-Free Global Optimization Algorithms under Limited Dimensions and Large Evaluation Budgets" published in IEEE Transactions on Evolutionary Computation [5]. In the paper, we performed a comprehensive computational comparison between stochastic and deterministic global optimization algorithms with twenty-five representative state-of-the-art methods selected from both classes. The experiments were set up with up to twenty dimensions and relatively large evaluation budgets (105 X n). Benchmarking was carried out in a significantly expanded version of the DIRECTGOLib v2.0 library, which included ten distinct collections of primarily continuous test functions. The evaluation of the methods focused on various aspects, such as solution quality, time complexity, and function evaluation usage. The rankings were determined using statistical tests and performance profiles. Our findings suggest that while state-of-the-art deterministic methods could find reasonable solutions with comparatively fewer function evaluations, most stochastic algorithms require more extensive evaluation budgets to deliver comparable results. However, the performance of stochastic algorithms excelled in more complex and higher-dimensional problems. These research findings offer valuable insights for practitioners and researchers, enabling them to tackle diverse optimization problems effectively.

This Hot Off the Press paper provides a brief summary of our recent work "Benchmarking Derivative-Free Global Optimization Algorithms under Limited Dimensions and Large Evaluation Budgets" published in IEEE Transactions on Evolutionary Computation [5]. In the paper, we performed a comprehensive computational comparison between stochastic and deterministic global optimization algorithms with twenty-five representative state-of-the-art methods selected from both classes. The experiments were set up with up to twenty dimensions and relatively large evaluation budgets (105 X n). Benchmarking was carried out in a significantly expanded version of the DIRECTGOLib v2.0 library, which included ten distinct collections of primarily continuous test functions. The evaluation of the methods focused on various aspects, such as solution quality, time complexity, and function evaluation usage. The rankings were determined using statistical tests and performance profiles. Our findings suggest that while state-of-the-art deterministic methods could find reasonable solutions with comparatively fewer function evaluations, most stochastic algorithms require more extensive evaluation budgets to deliver comparable results. However, the performance of stochastic algorithms excelled in more complex and higher-dimensional problems. These research findings offer valuable insights for practitioners and researchers, enabling them to tackle diverse optimization problems effectively.

Numerical benchmarking; derivative-free global optimization; evolutionary computation algorithms; nature-inspired meta-heuristics; deterministic algorithms

01.08.2024

Association for Computing Machinery, Inc

979-8-4007-0495-6

2024 Genetic and Evolutionary Computation Conference Companion, GECCO 2024 Companion

57–58

2