BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20251120T133321EST-8967W4bhna@132.216.98.100
DTSTAMP:20251120T183321Z
DESCRIPTION:Title: Efficient finite sample bounds via optimal transport and
  debiased ML without sample splitting”\n\nAbstract:\n\nFinite sample bound
 s are ubiquitous in statistics and machine learning\, underpinning applica
 tions ranging from multi-armed bandit problems to early stopping rules. Ho
 wever\, classical bounds are often overly conservative\, leading to subopt
 imal algorithms. In the first part of this talk\, I will propose a method 
 for deriving sharper bounds by bridging the gap between asymptotic limit t
 heorems and finite-sample concentration. We achieve this by exploiting rec
 ent advances in Optimal Transport\, Stein method and information theory. T
 he resulting bounds are efficient\, strictly valid in the finite-sample re
 gime\, and significantly tighter than the state-of-the-art. I will demonst
 rate how these bounds lead to direct algorithmic improvements.\n\nIn the s
 econd part of this talk\, we will study the generalized method of moments\
 , a key method for inference in causal inference. A recent line of work ha
 s shown how sample splitting in debiased machine learning enables the use 
 of generic machine learning estimators to estimate nuisance parameters whi
 le maintaining the asymptotic normality and root-n consistency of the targ
 et parameter. We show that when these auxiliary estimation algorithms sati
 sfy natural leave-one-out stability properties\, then sample splitting is 
 not required. This allows for sample re-use\, which can be beneficial in m
 oderately sized sample regimes.\n\n🔗 Zoom: https://mcgill.zoom.us/j/817340
 58047\n	Meeting ID: 817 3405 8047\n
DTSTART:20251121T183000Z
DTEND:20251121T193000Z
LOCATION:Room 1104\, Burnside Hall\, CA\, QC\, Montreal\, H3A 0B9\, 805 rue
  Sherbrooke Ouest
SUMMARY:Morgan Austern (Harvard University)
URL:/mathstat/channels/event/morgan-austern-harvard-un
 iversity-369094
END:VEVENT
END:VCALENDAR