Scholarlink

<p>The Bat-inspired algorithm (BA) is recognized as a prominent member of swarm intelligence algorithms, demonstrating significant potential in addressing global optimization challenges. This algorithm is distinguished by its unique feature of incorporating a local search mechanism, which enhances its convergence speed. However, an overemphasis on exploitation during the initial iterations can result in premature convergence, causing the algorithm to become trapped at a local optimum. To mitigate this limitation and enhance the exploration capabilities, this paper introduces an enhanced version of BA, termed complex-order BA (COBA). The evolution of position and velocity adaption laws in COBA, utilizing the concepts of complex-order derivative and the conjugate-order differential, broadens the memory associated with the prior behaviors of artificial bats and governs the algorithm’s convergence. First, we conducted a sensitivity analysis to identify the suitable parameters for the algorithm. Then, the performance of COBA is assessed using the well-known test functions derived from CEC 2017. Furthermore, the performance of the proposed algorithm is investigated for the weight optimization of feedforward neural networks and compared with state-of-the-art metaheuristic algorithms in the literature. COBA was also evaluated on three engineering design problems and consistently outperformed competing algorithms. It achieved superior results on 11 of 14 benchmark functions and showed strong performance in optimizing neural network weights, attaining the highest classification accuracy across all tested datasets. These findings highlight COBA’s reliability and effectiveness compared to state-of-the-art methods.</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

Complex-order Bat-inspired algorithm applied to neural network optimization for medical datasets

  • 期刊论文

  • 发表日期：

    2026年2月11日

查看全文
  • Ahmad Esfandiari,
  • Seyed Mehdi Abedi Pahnehkolaei,
  • Niki Nasiri

摘要

The Bat-inspired algorithm (BA) is recognized as a prominent member of swarm intelligence algorithms, demonstrating significant potential in addressing global optimization challenges. This algorithm is distinguished by its unique feature of incorporating a local search mechanism, which enhances its convergence speed. However, an overemphasis on exploitation during the initial iterations can result in premature convergence, causing the algorithm to become trapped at a local optimum. To mitigate this limitation and enhance the exploration capabilities, this paper introduces an enhanced version of BA, termed complex-order BA (COBA). The evolution of position and velocity adaption laws in COBA, utilizing the concepts of complex-order derivative and the conjugate-order differential, broadens the memory associated with the prior behaviors of artificial bats and governs the algorithm’s convergence. First, we conducted a sensitivity analysis to identify the suitable parameters for the algorithm. Then, the performance of COBA is assessed using the well-known test functions derived from CEC 2017. Furthermore, the performance of the proposed algorithm is investigated for the weight optimization of feedforward neural networks and compared with state-of-the-art metaheuristic algorithms in the literature. COBA was also evaluated on three engineering design problems and consistently outperformed competing algorithms. It achieved superior results on 11 of 14 benchmark functions and showed strong performance in optimizing neural network weights, attaining the highest classification accuracy across all tested datasets. These findings highlight COBA’s reliability and effectiveness compared to state-of-the-art methods.