BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:MARS via LASSO - Aditya Guntuboyina (UC Berkeley)
DTSTART:20211029T150000Z
DTEND:20211029T160000Z
UID:TALK162130@talks.cam.ac.uk
CONTACT:Qingyuan Zhao
DESCRIPTION:MARS is a popular method for nonparametric regression proposed
  by Friedman in 1991. MARS fits simple nonlinear and non-additive function
 s 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 co
 mbinations of functions in the MARS basis and putting a variation based co
 mplexity constraint. Our method is naturally connected to nonparametric fu
 nction estimation methods under smoothness constraints. Under natural desi
 gn assumptions\, we prove that our estimator achieves a rate of convergenc
 e that depends only logarithmically on dimension and thus avoids the usual
  curse of dimensionality to some extent. This is joint work with Dohyeong 
 Ki and Billy Fang. 
LOCATION:https://maths-cam-ac-uk.zoom.us/j/93998865836?pwd=VzVzN1VFQ0xjS3V
 DdlY0enBVckY5dz09
END:VEVENT
END:VCALENDAR