BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Multiscale Analysis of Bayesian CART - Veronika Rockova — Univer
 sity of Chicago
DTSTART:20191115T140000Z
DTEND:20191115T150000Z
UID:TALK130066@talks.cam.ac.uk
CONTACT:Dr Sergio Bacallado
DESCRIPTION:This paper affords new insights about Bayesian CART in the con
 text of structured wavelet shrinkage. We show that practically used Bayesi
 an CART priors lead to adaptive rate-minimax posterior concentration in th
 e supremum norm in Gaussian white noise\, performing optimally up to a log
 arithmic factor. To further explore the benefits of structured shrinkage\,
  we propose the g-prior for trees\, which departs from the typical wavelet
  product priors by harnessing correlation induced by the tree topology. Bu
 ilding on supremum norm adaptation\, an adaptive non-parametric Bernstein
 –von Mises theorem for Bayesian CART is derived using multi- scale techn
 iques. For the fundamental goal of uncertainty quantification\, we constru
 ct adaptive confidence bands with uniform coverage for the regression func
 tion under self-similarity. (Joint work with Ismael Castillo)
LOCATION:MR12
END:VEVENT
END:VCALENDAR