Title: Fast automatic smoothing in multiple generalized additive models
Authors: Yousra El Bachir - Ecole Polytechnique Fédérale de Lausanne (Switzerland) [presenting]
Anthony Davison - EPFL (Switzerland)
Abstract: A general statistical methodology for fitting distributions with parameters that depend smoothly on covariates through additive structures is presented. These multiple generalized additive models (GAMs) are estimated simultaneously and the optimal degree of penalization, which determines their smoothness, is incorporated automatically through a likelihood-based approach. The resulting method is statistically efficient and numerically stable while being simpler and faster than the gold standard, and can be extended to big-data settings easily.