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:20200308T070000
END:DAYLIGHT
BEGIN:STANDARD
TZNAME:EST
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
DTSTART:20191103T060000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
UID:682b25eb269ef
DTSTART;TZID=America/Toronto:20200917T160000
SEQUENCE:0
TRANSP:TRANSPARENT
DTEND;TZID=America/Toronto:20200917T160000
URL:/statistics-and-actuarial-science/events/department
 -seminar-neil-spencer-carnegie-mellon-university
SUMMARY:Department seminar by Neil Spencer\, Carnegie Mellon University
CLASS:PUBLIC
DESCRIPTION:Summary \n\nA NEW FRAMEWORK FOR MODELING SPARSE NETWORKS THAT M
 AKES SENSE (AND CAN\nACTUALLY BE FIT!)\n\nLatent position models are a ver
 satile tool when working with network\ndata. Applications include clusteri
 ng entities\, network visualization\,\nand controlling for unobserved caus
 al confounding. In traditional\ntreatments of the latent position model\, 
 the nodes’ latent positions\nare viewed as independent and identically d
 istributed random\nvariables. This assumption implies that the average nod
 e degree grows\nlinearly with the number of nodes in the network\, making 
 it\ninappropriate when the network is sparse. In the first part of this\nt
 alk\, I will propose an alternative assumption—that the latent\nposition
 s are generated according to a Poisson point process—and\nshow that it i
 s compatible with various levels of network sparsity. I\nwill also provide
  theory establishing that the nodes’ latent\npositions can be consistent
 ly estimated\, provided that the network\nisn't too sparse.  In the secon
 d part of the talk\, I will consider\nthe computational challenge of fitti
 ng latent position models to large\ndatasets. I will describe a new Markov
  chain Monte Carlo\nstrategy—based on a combination of split Hamiltonian
  Monte Carlo and\nFirefly Monte Carlo—that is much more efficient than t
 he standard\nMetropolis-within-Gibbs algorithm for inferring the latent po
 sitions.\nThroughout the talk\, I will use an advice-sharing network of\ne
 lementary school teachers within a school district as a running\nexample.\
 n\nPlease note: This talk will be hosted on Webex. To join please click\n
 on the following link: Department seminar by Neil Spencer\n[https://uwat
 erloo.webex.com/uwaterloo/onstage/g.php?MTID=e15ee553cf6b5ab678d359aa7d122
 a3ca].\n 
DTSTAMP:20250519T123659Z
END:VEVENT
END:VCALENDAR