BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Towards a better understanding of early stopping for boosting algo
 rithms - Yuting Wei\, Stanford University
DTSTART:20181102T160000Z
DTEND:20181102T170000Z
UID:TALK109666@talks.cam.ac.uk
CONTACT:Dr Sergio Bacallado
DESCRIPTION:In this talk\, I will discuss the behaviour of boosting algori
 thm for non-parametric regression. While non-parametric models offer great
  flexibility\, they can lead to overfitting and thus poor generalisation p
 erformance. For this reason\, procedures for fitting these models must inv
 olve some form of regularisation. Although early-stopping of iterative alg
 orithms is a widely-used form of regularisation in statistics and optimisa
 tion\, it is less well-understood than its analogue based on penalised reg
 ularisation. We exhibit a direct connection between a stopped iterate and 
 the localised Gaussian complexity of the associated function class which a
 llows us to derive explicit and optimal stopping rules. We will discuss su
 ch stopping rules in detail for various reproducing kernel Hilbert spaces\
 , and also extend these insights to broader classes of functions.
LOCATION:MR12
END:VEVENT
END:VCALENDAR