BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20251024T091811EDT-3238jKw4pr@132.216.98.100
DTSTAMP:20251024T131811Z
DESCRIPTION:Title: Adapting black-box machine learning methods for causal i
 nference.\n\nAbstract: I'll discuss the use of observational data to estim
 ate the causal effect of a treatment on an outcome. This task is complicat
 ed by the presence of 'confounders' that influence both treatment and outc
 ome\, inducing observed associations that are not causal. Causal estimatio
 n is achieved by adjusting for this confounding by using observed covariat
 e information. I'll discuss the case where we observe covariates that carr
 y sufficient information for the adjustment\, but where explicit models re
 lating treatment\, outcome\, covariates\, and confounding are not availabl
 e. For example\, in medical data the covariates might consist of a large n
 umber of convenience health measurements of which only an unknown subset a
 re relevant\, and even then in some totally unknown manner. Or\, the covar
 iates might be a passage of (natural language) text that describes the rel
 evant information. I'll describe an approach that adapts deep learning and
  embedding methods to produce representations of the covariate information
  targeted toward the causal adjustment problem. In particular\, I'll descr
 ibe how to modify standard architectures and training objectives to achiev
 e statistically efficient and practically useful causal estimates.\n
DTSTART:20200131T203000Z
DTEND:20200131T213000Z
LOCATION:Room 1104\, Burnside Hall\, CA\, QC\, Montreal\, H3A 0B9\, 805 rue
  Sherbrooke Ouest
SUMMARY:Victor Veitch (Columbia University)
URL:/mathstat/channels/event/victor-veitch-columbia-un
 iversity-303842
END:VEVENT
END:VCALENDAR