|
Activity Number:
|
558
- Semi- or Nonparametric Modeling for Data with Complex Structure
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 11, 2022 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Nonparametric Statistics
|
|
Abstract #320849
|
|
|
Title:
|
MARS via LASSO
|
|
Author(s):
|
Dohyeong Ki* and Billy Fang and Adityanand Guntuboyina
|
|
Companies:
|
University of California at Berkeley and Google LLC and University of California Berkeley
|
|
Keywords:
|
Constrained least squares estimation;
Hardy-Krause variation;
Infinite-dimensional optimization;
Locally adaptive regression spline;
Total variation regularization;
Trend filtering
|
|
Abstract:
|
MARS is a popular method for nonparametric regression introduced by Friedman in 1991. MARS fits simple nonlinear and non-additive functions to regression data. We propose and study a natural LASSO variant of the MARS method. Our method is based on least squares estimation over a convex class of functions obtained by considering infinite-dimensional linear combinations of functions in the MARS basis and imposing a variation based complexity constraint. We show that our estimator can be computed via finite-dimensional convex optimization and that it is naturally connected to nonparametric function estimation techniques based on smoothness constraints. Under a simple design assumption, we prove that our estimator achieves a rate of convergence that depends only logarithmically on dimension and thus avoids the usual curse of dimensionality to some extent. We implement our method with a cross-validation scheme for the selection of the involved tuning parameter and show that it has favorable performance compared to the usual MARS method in simulation and real data settings.
|
Authors who are presenting talks have a * after their name.
