BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Temporal Pointwise Convolutional Networks for Length of Stay Predi
 ction in the Intensive Care Unit - Emma Rocheteau (University of Cambridge
 )
DTSTART:20200519T121500Z
DTEND:20200519T131500Z
UID:TALK142378@talks.cam.ac.uk
CONTACT:Mateja Jamnik
DESCRIPTION:"ONLINE link":https://teams.microsoft.com/l/meetup-join/19%3a0
 29f32452e0a4506821c41ea466516e8%40thread.skype/1589449875820?context=%7b%2
 2Tid%22%3a%2249a50445-bdfa-4b79-ade3-547b4f3986e9%22%2c%22Oid%22%3a%22760a
 f26a-1349-4870-a967-af40fbad85e9%22%7d\n\nThe pressure of ever-increasing 
 patient demand and budget restrictions make hospital bed management a dail
 y challenge for clinical staff. Most critical\, is the efficient allocatio
 n of resource-heavy Intensive Care Unit (ICU) beds to the patients who nee
 d life support. Central to solving this problem is knowing for how long th
 e current set of ICU patients are likely to stay in the unit. In this work
  we propose a new deep learning model based on the combination of temporal
  convolution and pointwise (or 1x1) convolution\, to solve the length of s
 tay prediction task on the eICU critical care dataset. The model — which
  we refer to as Temporal Pointwise Convolution (TPC) — was developed usi
 ng a tailored\, domain-specific approach. We specifically design the model
  to mitigate for common challenges with Electronic Health Records\, such a
 s skewness\, irregular sampling and missing data. In doing so\, we have ac
 hieved significant performance benefits of 22-59% (metric dependent) over 
 the commonly used Long-Short Term Memory (LSTM) network.
LOCATION:Online on Teams
END:VEVENT
END:VCALENDAR