Data sets for the study "Automated Configuration of Genetic Algorithms by Tuning for Anytime Performance."
DOI10.5281/zenodo.4823492Zenodo4823492MaRDI QIDQ6690476FDOQ6690476
Dataset published at Zenodo repository.
Thomas Bäck, Hao Wang, Carola Doerr, Furong Ye
Publication date: 23 May 2021
Copyright license: Creative Commons Attribution 4.0 International
This is the result of the study Automated Configuration of Genetic Algorithms by Tuning for Anytime Performance. We compare grid search with three automated algorithm configuration methods, iterated racing (Irace), mixed-integer parallel efficient global optimization (MIP-EGO), and mixed-integer evolutionary strategies (MIES). The genetic algorithm (GA) is tuned for better the expected running time (ERT) and the area under the empirical cumulative distribution function curve (AUC). The result is tested on 25 pseudo-boolean problems. This Data set consists of 3parts: 1. dataand configurations:The performance of the configured GAs obtained by the configurators on 25 pseudo-Boolean problems defined inIOHprofiler(https://iohprofiler.github.io/), and the parameter settings of the GAs. 2. pvalues-le and pvalues-ge:The p-values of using WilcoxonMannWhitney two-sample rank-sum test to compare the runtimes to hit the target of the obtained GAs to the (1+1)-EA. The alternative hypothesis is that the evaluation times of (1+1)-EA is less than the evaluation times of the obtained GAs for pvalues-le.csv, and the alternative hypothesis is that the evaluation times of (1+1)-EA is greater than the evaluation times of the obtained GAs for pvalues-ge.csv, 3. Irace-OM and Irace-LO:The result of 20 Irace runs on Onemax and LeadingOnes. The algorithms are named as Irace-cost metric-Budget ratio t-ID of runs. The cost metric is ERT or AUC. The maximal evaluation budget of each configuration is \((0.5+0.1t) \times ERT_{(1+1)-EA}, t \in [0..15]\). Contact: if you have any questions or suggestions, please feel free to contactFurong Ye.
This page was built for dataset: Data sets for the study "Automated Configuration of Genetic Algorithms by Tuning for Anytime Performance."
