Automated Tailoring of Heuristic-Based Renyi’s Entropy Maximizers for Efficient Melanoma Segmentation^*

Heuristic-based optimizers, such as Metaheuristics (MHs), are powerful and versatile tools for solving complex optimization problems. However, selecting the most suitable MH for a specific task remains challenging due to the many available options. High-level approaches, like Hyper-Heuristics (HHs) within the Automated Algorithm Design and Configuration framework, offer a promising solution by tailoring MHs to meet the particular requirements of each problem. This work presents a novel approach for generating custom MHs using an HH approach that combines Search Operators (SOs) from well-established MHs to maximize Renyi’s entropy for efficient melanoma image segmentation. Our methodology automates the SO selection, resulting in a robust, metaphor-free MH design tailored for dermatoscopic image segmentation. Experimental results demonstrate that combining differential mutation and random flight operators achieves superior segmentation performance. This approach effectively reduces the computational burden and expert involvement, providing a great companion for skin cancer melanoma detection.