Multimodality in Multi-objective Optimization - More Boon than Bane?
EMO(2019)
摘要
This paper addresses multimodality of multi-objective (MO) optimization landscapes. Contrary to common perception of local optima, according to which they are hindering the progress of optimization algorithms, it will be shown that local efficient sets in a multi-objective setting can assist optimizers in finding global efficient sets. We use sophisticated visualization techniques, which rely on gradient field heatmaps, to highlight those insights into landscape characteristics. Finally, the MO local optimizer MOGSA is introduced, which exploits those observations by sliding down the multi-objective gradient hill and moving along the local efficient sets.
更多查看译文
关键词
Multi-objective optimization, Multimodality, Fitness landscapes, Basins of attraction, Local search, Gradients
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络