We are glad to announce the following seminar:
Thursday, 3rd of November 2022
12:00, Room AS02 (only in presence, no online)
Bocconi University, Via Roentgen 1, floor -1
Speaker: Clément Dombry (Université de Franche-Comté)
Title: Stone's theorem and minimax rate of convergence for distributional
regression
Abstract: Distributional regression aims at estimating the full conditional
distribution of the response variable given the covariates and is a
fundamental tool in probabilistic forecast. In this talk, we prove Stone's
theorem for distributional regression. It establishes universal consistency
in Wasserstein distance of a large class of models based on local weights.
Next we consider minimax rates of convergence on suitable classes of
distributions and prove that nearest neighbor and kernel algorithms are
rate optimal.
Best regards,
Giacomo Zanella