2024/08/19 by Amir Omeradzic, Omeradzic, Amir, Hans-Georg Beyer +1
Biochemistry, Genetics and Molecular Biology · #Evolution and Genetic Dynamics #FOS: Computer and information sciences #Gene expression and cancer classification #Neural and Evolutionary Computing (cs.NE)
paper · pdf · doi:10.48550/arxiv.2408.09761
openalex publication_date 2024/08/19 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
The mutation strength adaptation properties of a multi-recombinative (μ/μI, λ)-ES are studied for isotropic mutations. To this end, standard implementations of cumulative step-size adaptation (CSA) and mutative self-adaptation (σSA) are investigated experimentally and theoretically by assuming large population sizes (μ) in relation to the search space dimensionality (N). The adaptation is characterized in terms of the scale-invariant mutation strength on the sphere in relation to its maximum achievable value for positive progress. %The results show how the different σ-adaptation variants behave as μ and N are varied. Standard CSA-variants show notably different adaptation properties and progress rates on the sphere, becoming slower or faster as μ or N are varied. This is shown by investigating common choices for the cumulation and damping parameters. Standard σSA-variants (with default learning parameter settings) can achieve faster adaptation and larger progress rates compared to the CSA. However, it is shown how self-adaptation affects the progress rate levels negatively. Furthermore, differences regarding the adaptation and stability of σSA with log-normal and normal mutation sampling are elaborated.