University of Calabria, DIMES, 87036, Rende, (CS), Italy.
Lobachevsky State University, Institute of Information Technology, Mathematics and Mechanics, 603950, Nizhni Novgorod, Russia.
Sci Rep. 2018 Jan 11;8(1):453. doi: 10.1038/s41598-017-18940-4.
Global optimization problems where evaluation of the objective function is an expensive operation arise frequently in engineering, decision making, optimal control, etc. There exist two huge but almost completely disjoint communities (they have different journals, different conferences, different test functions, etc.) solving these problems: a broad community of practitioners using stochastic nature-inspired metaheuristics and people from academia studying deterministic mathematical programming methods. In order to bridge the gap between these communities we propose a visual technique for a systematic comparison of global optimization algorithms having different nature. Results of more than 800,000 runs on 800 randomly generated tests show that both stochastic nature-inspired metaheuristics and deterministic global optimization methods are competitive and surpass one another in dependence on the available budget of function evaluations.
在工程、决策、最优控制等领域，经常会遇到目标函数评估成本高昂的全局优化问题。有两个庞大但几乎完全不相交的社区（它们有不同的期刊、会议、测试函数等）致力于解决这些问题：一个广泛的实践者社区使用随机自然启发式元启发式算法，另一个来自学术界的人则研究确定性数学规划方法。为了弥合这两个社区之间的差距，我们提出了一种可视化技术，用于系统比较具有不同性质的全局优化算法。在 800 个随机生成的测试上进行的超过 80 万次运行的结果表明，随机自然启发式元启发式算法和确定性全局优化方法都是具有竞争力的，并且在可用的函数评估预算的依赖下相互超越。
