Akademik Veri Yönetim Sistemi
Applied Soft Computing, cilt.198, 2026 (SCI-Expanded, Scopus)
Multi-meme Memetic Algorithms are a class of self-adaptive approaches in memetic computing for tackling computationally hard optimisation problems. The memetic material representing algorithmic components, such as genetic operators and their settings to be used during the search process is co-evolved simultaneously along with the genetic material. The classical meme propagation method referred to as the simple inheritance mechanism selects the fitter parent's memes and transfers them to the offspring after crossover. This process represents the diffusion of ideas through the population, which actually mimics cultural evolution. There is almost no thorough analysis of other potential meme selection strategies for meme propagation in the scientific literature. In this study, we designed six different meme selection strategies and embedded them into the steady-state and trans-generational multimeme memetic algorithms, thus creating twelve different self-adaptive approaches. Using the HyFlex software library for cross-domain search, we conducted extensive experiments on ninety-eight instances from nine problem domains and observed consistently strong performance of the proposed strategies across varied population sizes and innovation rate settings. The results show that the proposed meme selection methods improve the cross-domain performance of the multimeme memetic algorithm in both steady-state and trans-generational regimes.