BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Interpretable and interactive machine learning - Dr Been Kim
DTSTART:20160729T100000Z
DTEND:20160729T110000Z
UID:TALK66923@talks.cam.ac.uk
CONTACT:Adrian Weller
DESCRIPTION:I envision a system that enables successful collaborations bet
 ween humans and machine learning models by harnessing the relative strengt
 h to accomplish what neither can do alone. Machine learning techniques and
  humans have skills that complement each other --- machine learning techni
 ques are good at computation on data at the lowest level of granularity\, 
 whereas people are better at abstracting knowledge from their experience\,
  and transferring the knowledge across domains. The goal of my research is
  to develop a framework for human-in-the-loop machine learning that enable
 s people to interact effectively with machine learning models to make bett
 er decisions using large datasets\, without requiring in-depth knowledge a
 bout machine learning techniques.\n \nIn this talk\, I present the Bayesia
 n Case Model (BCM)\, a general framework for Bayesian case-based reasoning
  (CBR) and prototype classification and clustering. BCM brings the intuiti
 ve power of CBR to a Bayesian generative framework. The BCM learns prototy
 pes\, the "quintessential" observations that best represent clusters in a 
 dataset\, by performing joint inference on cluster labels\, prototypes and
  important features. Simultaneously\, BCM pursues sparsity by learning sub
 spaces\, the sets of features that play important roles in the characteriz
 ation of the prototypes. The prototype and subspace representation provide
 s quantitative benefits in interpretability while preserving classificatio
 n accuracy. Human subject experiments verify statistically significant imp
 rovements to participants' understanding when using explanations produced 
 by BCM\, compared to those given by prior art. I demonstrate the applicati
 on of this model for an educational domain in which teachers cluster progr
 amming assignments to streamline the grading process.
LOCATION:CBL Room BE-438
END:VEVENT
END:VCALENDAR