Binary Tree Population Structure-Based Differential Evolution

For increasingly complex real-world optimization problems, a large number of researches have shown that differential evolution (DE) is an effective problem-solving algorithm. Despite its proven efficacy, DE still encounters challenges associated with rapid convergence and vulnerability to local optima. We introduce a novel methodology, namely binary tree population structure-based differential evolution (BTDE), to mitigate these issues. BTDE introduces a binary tree population structure comprising multiple layers of populations, strategically combined to exploit valuable information across diverse populations and ensure population diversity during the evolutionary process. To evaluate the performance of BTDE, we conduct rigorous assessments utilizing two benchmark test sets, including the CEC2017 benchmark functions and the CEC2011 real-world problems. Furthermore, we investigate the sensitivity of BTDE's parameters and conduct ablation studies to examine the individual layer's influence. The experimental results demonstrate that BTDE outperforms state-of-the-art DE variants, CEC competition winners, and selected powerful variants of the classical algorithm, substantiating its superiority in tackling complex problems.