Improving preference learning for MR-Sort using GPU

DA2PL: From Multiple-Criteria Decision Aid to Preference Learning, Nov 2022
Compiègne, France
Laurent Cabaret, Vincent Jacques, Vincent Mousseau
Texte intégral

@inproceedings{key,
  title = "Improving preference learning for MR-Sort using GPU",
  author = "Laurent Cabaret, Vincent Jacques, Vincent Mousseau",
  booktitle = "DA2PL: From Multiple-Criteria Decision Aid to Preference Learning",
  year = "2022",
  url = "https://da2pl.pre.utc.fr/wp-content/uploads/sites/70/2022/11/DA2PL_paper_370.pdf"
}

The Majority Rule Sorting (MR-Sort) method assigns alternatives evaluated on multiple criteria to one of the predefined ordered categories. The Inverse MR-Sort problem (Inv-MR-Sort) consists in computing MR-Sort parameters that match a dataset. Although Inv-MR-Sort is known to be computationally difficult, exact resolution approaches have been proposed in the literature, but are confronted to a computational barrier. In contrast, Sobrie et al. [12] tackled it with a heuristic. In this work, we aim at improving the computational efficiency of this heuristic approach by parallelization strategies using GPU.