A Novel Parametric benchmark generator for dynamic multimodal optimization
In most existing studies on dynamic multimodal optimization (DMMO), numerical simulations have been performed using the Moving Peaks Benchmark (MPB), which is a two-decade-old test suite that cannot simulate some critical aspects of DMMO problems. This study proposes the Deterministic Distortion and...
| Authors: | , , , , |
|---|---|
| Format: | article |
| Status: | Versión enviada para evaluación y publicación |
| Publication Date: | 2021 |
| Country: | España |
| Institution: | Basque Center for Applied Mathematics (BCAM) |
| Repository: | BIRD. BCAM's Institutional Repository Data |
| OAI Identifier: | oai:bird.bcamath.org:20.500.11824/1404 |
| Online Access: | http://hdl.handle.net/20.500.11824/1404 |
| Access Level: | Open access |
| Keyword: | Dynamic problem Evolutionary algorithm Global optimization Niching Performance indicator Robust peak ratio |
| Summary: | In most existing studies on dynamic multimodal optimization (DMMO), numerical simulations have been performed using the Moving Peaks Benchmark (MPB), which is a two-decade-old test suite that cannot simulate some critical aspects of DMMO problems. This study proposes the Deterministic Distortion and Rotation Benchmark (DDRB), a method to generate deterministic DMMO test problems that can simulate more diverse types of challenges when compared to existing benchmark generators for DMMO. DDRB allows for controlling the intensity of each type of challenge independently, enabling the user to pinpoint the pros and cons of a DMMO method. DDRB first develops an existing approach for generation of static multimodal functions in which the difficulty of global optimization can be controlled. Then, it proposes a scaling function to dynamically change the relative distribution, shapes, and sizes of the basins. A deterministic technique to control the regularity of the pattern in the change is also proposed. Using these components, a parametric test suite consisting of ten test problems is developed for DMMO. Mean Robust Peak Ratio for measuring the performance of DMMO methods is formulated to overcome the sensitivity of the conventional peak ratio indicator to the predefined threshold and niche radius. Numerical results of a successful multimodal optimization method, when augmented with a simple strategy to utilize previous information, are provided on the proposed test problems in different scenarios with the aim of serving as a reference for future studies. |
|---|