Evolutionary Multi-Mode Slime Mould Optimization: A Hyper-Heuristic Algorithm Inspired by Slime Mould Foraging Behaviors

In this paper, we propose a novel hyper-heuristics algorithm named Evolutionary Multi-mode Slime mould Op-timization (EMSMO) inspired by the slime mould foraging behaviors for continuous optimization. The EMSMO is a com-bination of High-Level and Low-Level components. In the Low-Level component, we formulate the foraging behaviors of slime mould and design four generation strategies: Search for food, Approach food, Wrap food, and Re-initialization. In the High-Level component, we design an improvement-based probabilistic selection function containing the probability of improvement (P I) and the normalized improvement (N I) to determine the sequence of heuristics. Considering that a certain strategy may have different performances in various optimization stages, we adopt a sliding window to emphasize the recent optimization information to the construction of the optimization sequence. To evaluate the performance of our proposal, we run EMSMO with 6 classic or advanced Evolutionary Algorithms (EAs) on CEC2017 benchmark functions with 30 independent trial runs and compare them fairly. The statistical results show that our proposed EMSMO is competitive with compared methods and has great potential to deal with continuous optimization problems.