Cumulative Step Size Adaptation for Adaptive SEMO in Integer Space (Code and Dataset)

DOI10.5281/zenodo.14219908Zenodo14219908MaRDI QIDQ6710033FDOQ6710033

Dataset published at Zenodo repository.

Markus Wagner, Günter Rudolph

Publication date: 26 November 2024

Copyright license: MIT license

Description

This is the code and the dataset for our paper published at EMO 2025, https://emo2025.org Title: "Cumulative Step Size Adaptation for AdaptiveSEMO in Integer Space"Abstract: Parameter control involves dynamically adjusting the parameter values of the evolutionary algorithm throughout the optimization process, including parameters like mutation rate and operator selection.Self-adaptation can improve the performance and robustness of the algorithm, however, parameter control mechanisms themselves need to be designed and configured carefully.In this article, we review the cumulative step size adaptation method originally proposed for single-objective optimization over continuous variables and recast it for deployment in multiobjective optimization over unbounded integer space.We contribute a systematic investigation of its hyperparameters which shows that while (1) the very best configurations remain problem-specific, (2) the performance of the algorithm is largely independent of the self-adaptation scheme's parameterization and initial configuration.

This page was built for dataset: Cumulative Step Size Adaptation for Adaptive SEMO in Integer Space (Code and Dataset)