BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Learning with limited supervision - Stefano Ermon\, Stanford
DTSTART:20161201T110000Z
DTEND:20161201T120000Z
UID:TALK67587@talks.cam.ac.uk
CONTACT:Adrian Weller
DESCRIPTION:Many of the recent successes of machine learning have been cha
 racterized by the availability of large quantities of labeled data. Noneth
 eless\, we observe that humans are often able to learn with very few label
 ed examples or with only high level instructions for how a task should be 
 performed. In this talk\, I will present some new approaches for learning 
 useful models in contexts where labeled training data is scarce or not ava
 ilable at all. 1) I will introduce new techniques for learning generative 
 models\, including the use of random projections to simplify probabilistic
  models while preserving most of the information\, and a new boosting fram
 ework to learn ensembles of models. 2) I will discuss ways to use prior kn
 owledge (such as physical laws) to provide supervision\, showing how we ca
 n learn to solve useful tasks\, including object tracking\, without any la
 beled data. 3) Finally\, I will introduce new approaches to leverage spati
 o-temporal structure in semi-supervised learning frameworks. I will presen
 t applications of these ideas to address development and sustainability is
 sues\, including new scalable methods to map poverty and monitor food secu
 rity in developing countries using satellite imagery.\n\nBio: Stefano Ermo
 n is an Assistant Professor in the Department of Computer Science at Stanf
 ord University\, where he is affiliated with the Artificial Intelligence L
 aboratory and the Woods Institute for the Environment. He completed his Ph
 D in computer science at Cornell in 2015. His research interests include t
 echniques for scalable and accurate inference in graphical models\, statis
 tical modeling of data\, large-scale combinatorial optimization\, and robu
 st decision making under uncertainty\, and is motivated by a range of appl
 ications\, in particular ones in the emerging field of computational susta
 inability. Stefano has won several awards\, including two Best Student Pap
 er Awards\, one Runner-Up Prize\, and a McMullen Fellowship.
LOCATION:CBL Room BE-438
END:VEVENT
END:VCALENDAR