BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Intelligible Machine Learning Models for HealthCare - Rich Caruana
 \, MSR Redmond
DTSTART:20150603T100000Z
DTEND:20150603T110000Z
UID:TALK59579@talks.cam.ac.uk
CONTACT:Microsoft Research Cambridge Talks Admins
DESCRIPTION:In machine learning often a tradeoff must be made between accu
 racy and intelligibility: the most accurate models usually are not very in
 telligible (e.g.\, random forests\, boosted trees\, and neural nets)\, and
  the most intelligible models usually are less accurate (e.g.\, linear or 
 logistic regression).  This tradeoff often limits the accuracy of models t
 hat can be applied in mission-critical applications such as healthcare whe
 re being able to understand\, validate\, edit\, and trust a learned model 
 is important. We have developed a learning method based on generalized add
 itive models (GAMs) that is often as accurate as full complexity models\, 
 but remains as intelligible as linear/logistic regression models.  In the 
 talk I’ll present two case studies where these high-performance generali
 zed additive models (GA2Ms) are applied to healthcare problems yielding in
 telligible models with state-of-the-art accuracy.  In the pneumonia risk p
 rediction case study\, the intelligible model uncovers surprising patterns
  in the data that previously had prevented complex learned models from goi
 ng to clinical trial\, but because it is intelligible and modular allows t
 hese patterns to easily be recognized and removed. In the 30-day hospital 
 readmission case study\, we show that the same methods scale to large data
 sets containing hundreds of thousands of patients and thousands of attribu
 tes while remaining intelligible and providing accuracy comparable to the 
 best (unintelligible) machine learning methods.
LOCATION:Auditorium\, Microsoft Research Ltd\, 21 Station Road\, Cambridge
 \, CB1 2FB
END:VEVENT
END:VCALENDAR