Sparse zero-sum games as stable functional feature selection

Sparse zero-sum games as stable functional feature selection Sokolovska N. auteur aut IRD Teytaud O. auteur aut IRD Rizkalla S. auteur aut IRD Consortium M. auteur aut IRD Clement K. auteur aut IRD Zucker Jean-Daniel auteur aut IRD text journalArticle eng text/pdf reformatted digital access In large-scale systems biology applications, features are structured in hidden functional categories whose predictive power is identical. Feature selection, therefore, can lead not only to a problem with a reduced dimensionality, but also reveal some knowledge on functional classes of variables. In this contribution, we propose a framework based on a sparse zero-sum game which performs a stable functional feature selection. In particular, the approach is based on feature subsets ranking by a thresholding stochastic bandit. We provide a theoretical analysis of the introduced algorithm. We illustrate by experiments on both synthetic and real complex data that the proposed method is competitive from the predictive and stability viewpoints. specialized 020 Plos One 10 9 e0134683 [16 p.] 2015 1932-6203 https://www.documentation.ird.fr/hor/fdi:010065259 10.1371/journal.pone.0134683 1932-6203 [F B010065259] https://www.documentation.ird.fr/hor/fdi:010065259 https://horizon.documentation.ird.fr/exl-doc/pleins_textes/divers17-08/010065259.pdf IRD - Base Horizon / Pleins textes 2015-10-07 2017-09-06 fdi:010065259 fre