of which were deemed not relevant for our purpose of AUD classification. The method adapted in the current analysis is based on the Lasso method, as implemented in Fonti and Belitser [84]. The maximum number of output features “pmax” was set to 10% (i.e., 56 of the total 561 variables). A 10-fold cross-validation and lambda thresholding with 1 SE (λ1se) were set to extract the final set of key variables. The area under the curve (AUC) was plotted to determine the classification performance of the selected features. The final subset resulting from the feature selection process included 21 rsFC variables (see Table A1, Appendix A).
