This dataset contains (1) all code needed to reproduce our results, (2) the 20 data sets on which we report results, and (3) all experiment results, related to the paper: "Boosting Revisited: Benchmarking and Advancing LP-Based Ensemble Methods" as published in TMLR, see: https://openreview.net/forum?id=lscC4PZUE4.
The code is written in Python 3.12, a README inside the zipped code folder provides more details on setting up and running the code.
In the zipped data sets folder, we provide a README with more information on our preprocessing steps and links to the orginal sources from which we retrieved the data sets.
Each experiment is outputted to a JSON file. We include both the results as reported in the paper (the best-found hyperparameter setting) and all experiments related to other hyperparameter settings. The JSON files are organized in zipped folders per experiment type. See the README for further details.
" authors: - family-names: Akkerman given-names: Fabian orcid: "https://orcid.org/0000-0001-8055-9864" - family-names: Ferry given-names: Julien orcid: "https://orcid.org/0000-0002-8764-0080" - family-names: Artigues given-names: Christian orcid: "https://orcid.org/0000-0002-9766-9864" - family-names: Hebrard given-names: Emmanuel orcid: "https://orcid.org/0000-0003-3131-0709" - family-names: Vidal given-names: Thibaut orcid: "https://orcid.org/0000-0001-5183-8485" title: "Code, Data, and Experimental Results for "Boosting Revisited: Benchmarking and Advancing LP-Based Ensemble Methods"" keywords: version: 1 identifiers: - type: doi value: 10.4121/f82dcdaa-fc94-43c5-b66d-02579bd3de4f.v1 license: MIT date-released: 2025-10-22