BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Drupal iCal API//EN
X-WR-CALNAME:Events items teaser
X-WR-TIMEZONE:America/Toronto
BEGIN:VTIMEZONE
TZID:America/Toronto
X-LIC-LOCATION:America/Toronto
BEGIN:DAYLIGHT
TZNAME:EDT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
DTSTART:20190310T070000
END:DAYLIGHT
BEGIN:STANDARD
TZNAME:EST
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
DTSTART:20191103T060000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
UID:6829f865c7d33
DTSTART;TZID=America/Toronto:20200130T100000
SEQUENCE:0
TRANSP:TRANSPARENT
DTEND;TZID=America/Toronto:20200130T100000
URL:/statistics-and-actuarial-science/events/department
 -seminar-hyukjun-jay-gweon-western-university
SUMMARY:Department seminar by Hyukjun (Jay) Gweon\, Western University
CLASS:PUBLIC
DESCRIPTION:Summary \n\nBATCH-MODE ACTIVE LEARNING FOR REGRESSION AND ITS A
 PPLICATION TO THE\nVALUATION OF LARGE VARIABLE ANNUITY PORTFOLIOS\n\nSuper
 vised learning algorithms require a sufficient amount of labeled\ndata to 
 construct an accurate predictive model. In practice\,\ncollecting labeled 
 data may be extremely time-consuming while\nunlabeled data can be easily a
 ccessed. In a situation where labeled\ndata are insufficient for a predict
 ion model to perform well and the\nbudget for an additional data collectio
 n is limited\, it is important\nto effectively select objects to be labele
 d based on whether they\ncontribute to a great improvement in the model's 
 performance. In this\ntalk\, I will focus on the idea of active learning t
 hat aims to train\nan accurate prediction model with minimum labeling cost
 . In\nparticular\, I will present batch-mode active learning for regressio
 n\nproblems. Based on random forest\, I will propose two effective random\
 nsampling algorithms that consider the prediction ambiguities and\ndiversi
 ties of unlabeled objects as measures of their informativeness.\nEmpirical
  results on an insurance data set demonstrate the\neffectiveness of the pr
 oposed approaches in valuing large variable\nannuity portfolios (which is 
 a practical problem in the actuarial\nfield). Additionally\, comparisons w
 ith the existing framework that\nrelies on a sequential combination of uns
 upervised and supervised\nlearning algorithms are also investigated.\n 
DTSTAMP:20250518T151029Z
END:VEVENT
END:VCALENDAR