Scholarlink

<p>This paper introduces an efficient variant of Competitive Differential Evolution (CDE), named Parameter Adaptive CDE with Dynamic Local Search (PaCDE-DLS), to overcome the limitations of DE, including high-level parameter dependency and problem structure dependency. Two primary components are integrated into PaCDE-DLS to enhance the performance: The Success History Adaptation (SHA) mechanism adjusts control parameters based on past successful evolutionary trajectories, and the Dynamic Local Search (DLS) operator refines candidate solutions in the later optimization stage to achieve a higher convergence accuracy. We implement rigorous experiments in CEC2017, CEC2020, CEC2022, and seven engineering problems to benchmark PaCDE-DLS against ten state-of-the-art optimizers. Additionally, ablation studies further highlight the independent contributions of the SHA mechanism and DLS operator. Moreover, We extend PaCDE-DLS to an ensemble learning framework for potato plant disease diagnosis, where top-3 high-accuracy deep learning models are fused using a PaCDE-DLS-optimized soft voting scheme. Experimental results demonstrate that PaCDE-DLS-Ensemble performed well and achieved an accuracy of 99.768%, precision of 99.772%, recall of 99.768%, and F1 score of 99.767%, which underscores the broad applicability of PaCDE-DLS in real-world scenarios. The source code of this research can be downloaded from <a href="https://github.com/RuiZhong961230/PaCDE-DLS">https://github.com/RuiZhong961230/PaCDE-DLS</a>.</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

Parameter adaptive competitive differential evolution with local search

  • 期刊论文

  • 发表日期：

    2026年4月7日

查看全文
  • Rui Zhong,
  • Zhongmin Wang,
  • Yaning Xiao,
  • Yujun Zhang,
  • Junbo Jacob Lian,
  • Jun Yu,
  • Zhennao Cai,
  • Zhiyong Pan,
  • Huiling Chen,
  • Sudan Yu

摘要

This paper introduces an efficient variant of Competitive Differential Evolution (CDE), named Parameter Adaptive CDE with Dynamic Local Search (PaCDE-DLS), to overcome the limitations of DE, including high-level parameter dependency and problem structure dependency. Two primary components are integrated into PaCDE-DLS to enhance the performance: The Success History Adaptation (SHA) mechanism adjusts control parameters based on past successful evolutionary trajectories, and the Dynamic Local Search (DLS) operator refines candidate solutions in the later optimization stage to achieve a higher convergence accuracy. We implement rigorous experiments in CEC2017, CEC2020, CEC2022, and seven engineering problems to benchmark PaCDE-DLS against ten state-of-the-art optimizers. Additionally, ablation studies further highlight the independent contributions of the SHA mechanism and DLS operator. Moreover, We extend PaCDE-DLS to an ensemble learning framework for potato plant disease diagnosis, where top-3 high-accuracy deep learning models are fused using a PaCDE-DLS-optimized soft voting scheme. Experimental results demonstrate that PaCDE-DLS-Ensemble performed well and achieved an accuracy of 99.768%, precision of 99.772%, recall of 99.768%, and F1 score of 99.767%, which underscores the broad applicability of PaCDE-DLS in real-world scenarios. The source code of this research can be downloaded from https://github.com/RuiZhong961230/PaCDE-DLS.