Multi-swarm UPSO algorithm based on seed strategy for atomic clusters structure optimization. (December 2021)

• Record Type: Journal Article
• Authors: Tang, Xinghua; Liu, Jing; Zhu, Jingjing; Zhou, Lihai; Zhang, Yining
• Is Part Of: Computational biology and chemistry. Volume 95(2021)
• Publication Date: 2021-12
• Subjects: Atomic clusters -- Multi-swarm mechanism -- Particle swarm optimization -- Seed strategy -- Unified particle swarm optimization -- Information communication strategy; Chemistry -- Data processing -- Periodicals; Biology -- Data processing -- Periodicals; Biochemistry -- Data processing; Biology -- Data processing; Molecular biology -- Data processing; Periodicals; Electronic journals; 542.85
• DOI: 10.1016/j.compbiolchem.2021.107598
• Languages: English
• ISSNs: 1476-9271

Abstract

Abstract: Particle Swarm Optimization (PSO) algorithm is prone to get trapped in local optima and insufficient information exchange among particles. To solve this problem, this paper proposes a Multi-swarm Unified Particle Swarm Optimization algorithm based on Seed Strategy (SS-DMS-UPSO) to optimize the atomic clusters structure. In this algorithm, the population is divided into some sub-populations evolving randomly and evenly, and each sub-population uses UPSO algorithm with different unification factors to evolve independently in parallel. After a certain number of independent evolution, the particles of all sub-populations are merged into a new population, and the population is again randomly divided into average sub-populations. Iterate the algorithm repeatedly in this way. And finally the global best particle can be obtained. The experimental results show that the SS-DMS-UPSO algorithm can search for the optimal structure or extremely similar optimal structure for atomic clusters with atomic numbers between 2 and 31. For atomic clusters with atomic numbers between 32 and 35, the algorithm can find its approximate optimal structure. Compared with other algorithms, the difference between the lowest energy value and the ideal energy value obtained by the SS-DMS-UPSO algorithm is much smaller. It means that its optimal structure of the atomic clusters is closer to the stable structure, and the algorithm is more stable, which proves the effectiveness of the SS-DMS-UPSO algorithm. Graphical abstract: ga1 Highlights: Seed strategy can improve the probability of reaching global convergence Seed strategy enables to maintain algorithm stability Information Communication Strategy promotes particles information exchange