Gravitational search algorithm: a comprehensive analysis of recent variants

• Published in: Multimedia tools and applications Vol. 80; no. 5; pp. 7581 - 7608
• Main Authors: Mittal, Himanshu, Tripathi, Ashish, Pandey, Avinash Chandra, Pal, Raju
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.02.2021; Springer Nature B.V
• Subjects: Algorithms; Breast cancer; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Experiments; Gravitation; Multimedia Information Systems; Optimization; Parameter modification; Performance evaluation; Qualitative analysis; Search algorithms; Segmentation; Special Purpose and Application-Based Systems; Optimization algorithm; Gravitational search algorithm; Nature-inspired algorithm
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-020-09831-4

Gravitational search algorithm is a nature-inspired algorithm based on the mathematical modelling of the Newton’s law of gravity and motion. In a decade, researchers have presented many variants of gravitational search algorithm by modifying its parameters to efficiently solve complex optimization problems. This paper conducts a comparative analysis among ten variants of gravitational search algorithm which modify three parameters, namely K b e s t , velocity, and position. Experiments are conducted on two sets of benchmark categories, namely standard functions and CEC2015 functions, including problems belonging to different categories such as unimodal, multimodal, and unconstrained optimization functions. The performance comparison is evaluated and statistically validated in terms of mean fitness value and convergence graph. In experiments, IGSA has achieved better precision with balanced trade-off between exploration and exploitation. Moreover, triple negative breast cancer dataset has been considered to analysis the performance of GSA variants for the nuclei segmentation. The variants performance has been analysed in terms of both qualitative and quantitive with aggregated Jaccard index as performance measure. Experiments affirm that IGSA-based method has outperformed other methods.