A New Explicit Penalty Method for Evolutionary Multimodal Optimization
提出一种显式惩罚方法，通过量化解的多样性并整合目标值，平衡多样性与目标值，有效定位多个最优解，实验表明优于多种流行和最新算法。
When employing evolutionary algorithms (EAs) to solve multimodal optimization problems (MMOPs), effectively utilizing diversity information is crucial to prevent the population from converging to a single peak. This requires balancing diversity and objective value—a challenge that inherently constitutes a penalty problem. Although some implicit penalty methods have been proposed to address this issue, most lack flexibility in penalty formulation. In this study, we present a novel explicit penalty method (EPM) designed to effectively leverage diversity information for multimodal optimization. First, the diversity of a solution is quantified by its distance to the nearest neighbor with a better objective value. Then, an explicit penalty function is formulated by integrating diversity and objective value. This function facilitates the capture of multiple peaks and balances the search among them. If a reasonable number of peaks are identified, a local search is applied to each for refinement; otherwise, a global search is conducted across the decision space. Through this adaptive process, EPM locates multiple optima both efficiently and accurately. Extensive experiments demonstrate that EPM outperforms several multimodal optimization methods, including 11 popular approaches, eight recent state-of-the-art algorithms, and an IEEE CEC competition winner. Moreover, even when integrated with classic differential evolution (DE), EPM exhibits highly competitive performance.