BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Machine Learning for Critical Care: Using Neural Networks to Predi
 ct Patient Mortality - Conor Cafolla\, University of Cambridge
DTSTART:20201028T143000Z
DTEND:20201028T150000Z
UID:TALK151072@talks.cam.ac.uk
CONTACT:Lisa Masters
DESCRIPTION:Intensive care units are constantly under strain\, with many v
 ulnerable people needing urgent critical care.  Often doctors do not have 
 the capacity to accommodate everyone\, and so it is important that patient
 s are discharged when it is safe to do so.  However\, discharging a patien
 t at the wrong time may result in the patient being readmitted or simply n
 ot surviving\;  both of these cases are highly undesirable.  Therefore\, i
 t is useful for there to be an understanding of what factors contribute mo
 st to the mortality rate of patients in intensive care units.\n\nIn the pr
 esent work\, machine learning models are trained to predict whether a give
 n patient with certain measurements will live or die.  Neural networks wit
 h one hidden layer and two outcomes (alive or deceased) are used in these 
 models\, and local minima of the neural network loss functions are found u
 sing basin-hopping methods implemented with the open source software GMIN.
   Area Under the Curve (AUC) values were used to evaluate these models\, a
 nd AUC values above 0.8 were found for models involving Glasgow Coma Scale
  (GCS) scores and Blood Urea Nitrogen (BUN) measurements.  The effect of u
 sing a model trained on one time window and evaluated on different time wi
 ndows is also investigated\, and we find that the AUC value decreases but 
 not substantially.  Finally\, the effect of label noise is investigated\, 
 and it is found that neural networks used on noisy iris flower data are ro
 bust\, giving AUC values close to 1\, even for moderately high proportions
  of mislabelled training data.\n\nThe project is heading in a direction fo
 cusing on the thermodynamic properties of the energy landscapes defined by
  neural network loss functions.  By identifying the abstract analogue of `
 `phase transitions" for this landscape\, using some fictitious temperature
 \, it is hoped that complementary neural network solutions on the landscap
 e may be found and combined to yield overall better solutions.\n
LOCATION:Zoom - Meeting ID: 924 3189 5042 Passcode: 306870
END:VEVENT
END:VCALENDAR