Elastic Net avec gestion des interactions et débiaisage
In EGC 2021, vol. RNTI-E-37, pp.269-276
We present some statisticals results for a penalized and de-biased regression estimator, to handle in high dimension interactions and sparsity. The statistical analysis focuses on the approach proposed in (Bascou et al., 2020), in particular to illustrate the functioning of debiasing. Moreover, it is shown that it allows the selection of fewer variables than the model without debiaising on simulated and real data.