[DATA SUBMISSION] Benchmarking CMA-ES with Basic Integer Handling on a Mixed-Integer Test Problem Suite

[trmarty]

Reference

Tristan Marty, Yann Semet, Anne Auger, Sébastion Héron, and Nikolaus Hansen. 2023. Benchmarking CMA-ES with Basic Integer Handling on a Mixed-Integer Test Problem Suite. In Proceedings of Genetic and Evolutionary Computation Conference Companion, Lisbon, Portugal, July 15–19, 2023 (GECCO ’23 Companion), 8 pages. DOI: 10.1145/3583133.3596411

Description of the Algorithm(s)

The algorithms benchmarked are some variant of CMA-ES for handling mixed variable on the bbob-mixint test suite.

• cma-IH : pycma implementation of CMA-ES version 3.3 WITH the default integer handling option.
• cma-NoIH : pycma implementation of CMA-ES version 3.3 WITHOUT the default integer handling option.
• cmaIH1e-1 : pycma implementation of CMA-ES version 3.3 with a larger coordinate-wise lower bound on the sample standard deviatioon of $\sigma_{LB}=10^{-1}$

Link to Data

• cma-IH : https://github.com/trmarty/MixedIntegerData/blob/main/raw_data/cma-IH.zip
• cma-NoIH : https://github.com/trmarty/MixedIntegerData/blob/main/raw_data/cma-NoIH.zip
• cmaIH1e-1 : https://github.com/trmarty/MixedIntegerData/blob/main/raw_data/cmaIH1e-1.zip

[brockho]

Dear Tristan,

Many thanks for your data submission (and sorry for having been quiet for so long). I realized that your data set for cma-IH misses some data in (the non-mandatory) dimension 160 - if I am not mistaken, it is instance 8 for f23 and instances 8 and 9 for f24 that are missing. Would you be so kind and send us these data (here or per email/skype/jitsi/...) such that we can put the data sets to the official COCO data archive? Many thanks in advance.