Regarding the variable selection, we set the estimation metric to the MAD value , as advised in. To reduce the computational effort, we adopted the simpler 2/3 and 1/3 holdout split as the internal validation method. The sensitivity analysis parameter was set to L = 6, i.e. xa ∈ {−1.0, −0.6, . . . , 1.0} for a standardized input. As a reasonable balance between the pressure towards simpler models and the increase of computational search, the stopping criterion was set to 2 iterations without any improvement or when only one input is available.