[Colloquium] Talks at TTIC: Hossein Azari, Harvard University -- Thurs. May 1st!

[Dawn Ellis]

When:     Thursday, May 1st at Noon

Where:    TTIC, 6045 S Kenwood Avenue, 5th Floor, Room #526

Who:       Hossein Azari, Harvard University

Title:        Revisiting Random Utility Models for Rank Data

Abstract:

Random utility models (RUMs) are used to model rank data in many
applications, including rank aggregation, computational social choice
and econometrics. A real-valued score is associated with each
alternative, typically independently sampled according to a
parameterized distribution. Alternatives are ranked by score. The
Plackett-Luce model is a special case for which tractable maximum
likelihood estimators are available. In this talk, I describe
generalized RUMs that capture heterogeneity of agents through more
flexible distributions, mixture models, and through non-parametric
models that also capture dependence across alternatives. I provide an
algorithm for maximum likelihood estimation, along with a fast,
rank-breaking estimation procedure using generalized method of
moments. Empirical results show that the new methods perform better
using various fitness metrics and prediction power in rank completion
for recommender systems.

Bio:

Hossein Azari is a PhD candidate in Computer Science at Harvard
University. He received his B.S. in Electrical Engineering from Sharif
University (with distinguished student award) and A.M. in Statistics
from Harvard University. His research focuses on statistical machine
learning with applications to social computation, economics and social
networks. He is a Siebel scholar class of 2014.

Hosts:  Nati Srebro, nati at ttic.edu  and  Ryoto Tomioka, tomioka at ttic.edu

-- 
*Dawn Ellis*
Administrative Coordinator,
Bookkeeper
773-834-1757
dellis at ttic.edu

TTIC
6045 S. Kenwood Ave.
Chicago, IL. 60637
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.cs.uchicago.edu/pipermail/colloquium/attachments/20140429/8f5103b3/attachment.htm