Scholarlink

<p>During the last decade, various nature-inspired metaheuristic algorithms have been proposed to tackle and address the optimization problems in various fields. Among them, the Starfish Optimization Algorithm (SFOA) is a recent population-based optimization that simulates starfish behavior in nature. SFOA showed efficient performance and outperformed 100 optimizers for solving optimization and engineering problems. However, the computational time of SFOA increases when solving large-scale problems. In this paper, we proposed island-based SFOA using the Apache Spark framework, called IS-SFOA. The main idea is to divide the population into subpopulations (islands). Each island evolves independently and exchanges the information at periodic intervals. We conducted comprehensive experiments to evaluate the robustness, efficiency, and scalability of IS-SFOA on benchmark functions and three constrained engineering optimization problems. For solution quality, IS-SFOA was compared with the serial SFOA baseline and two recent Spark-based metaheuristics, Spark-SCA and Spark-AOA. The results show that IS-SFOA preserves the optimization efficacy of serial SFOA. Moreover, IS-SFOA achieves superior performance and outperforms Spark-SCA and Spark-AOA. For scalability, IS-SFOA was evaluated under increasing workloads with population sizes up to 120,000 and dimensionalities up to 1,000, demonstrating substantial and stable speedup as the number of cluster nodes increases and effective utilization of distributed computing resources.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

The island-based Spark implementation of the Starfish Optimization Algorithm

  • 期刊论文

  • 发表日期：

    2026年6月23日

查看全文
  • Jamil Al-Sawwa

摘要

During the last decade, various nature-inspired metaheuristic algorithms have been proposed to tackle and address the optimization problems in various fields. Among them, the Starfish Optimization Algorithm (SFOA) is a recent population-based optimization that simulates starfish behavior in nature. SFOA showed efficient performance and outperformed 100 optimizers for solving optimization and engineering problems. However, the computational time of SFOA increases when solving large-scale problems. In this paper, we proposed island-based SFOA using the Apache Spark framework, called IS-SFOA. The main idea is to divide the population into subpopulations (islands). Each island evolves independently and exchanges the information at periodic intervals. We conducted comprehensive experiments to evaluate the robustness, efficiency, and scalability of IS-SFOA on benchmark functions and three constrained engineering optimization problems. For solution quality, IS-SFOA was compared with the serial SFOA baseline and two recent Spark-based metaheuristics, Spark-SCA and Spark-AOA. The results show that IS-SFOA preserves the optimization efficacy of serial SFOA. Moreover, IS-SFOA achieves superior performance and outperforms Spark-SCA and Spark-AOA. For scalability, IS-SFOA was evaluated under increasing workloads with population sizes up to 120,000 and dimensionalities up to 1,000, demonstrating substantial and stable speedup as the number of cluster nodes increases and effective utilization of distributed computing resources.