[Colloq] CCIS Colloquium

[panarese at ccs.neu.edu]

The College of Computer and Information Science presents: 



CCIS Colloquium 





Title: Algorithmic Recommender Systems 

Speaker: Boaz Patt-Shamir, Tel Aviv University 

When: Friday, April 29, 2011, 11 AM 

Where: 366 WVH 

Abstract: 

Recommender systems help users identify objects they may find 
interesting, where objects may be books to read, films to watch, web 
pages to browse, and even other users to contact. Formally, the input 
to the system is the known preferences of the users, as deduced 
somehow from their past choices.  While many interesting ideas have 
been developed to analyze characteristics of users (or objects) based 
on past choices, this approach suffers from a foundational theoretical 
flaw: feedback is ignored. Put simply, the setting is such that 
choices determine recommendations, but recommendations are supposed to 
influence choices. In a recent line of work this gap was bridged by a 
simple algorithmic model that assumes that the system may ask the 
user's opinion on any object, and not only make recommendations about 
supposedly `nice' objects. Typically, algorithms in this model ask 
users for their opinions on controversial objects, and in return, the 
output consists of almost complete reconstruction of user 
preferences. In this talk we discuss this model and survey some basic 
and recent results. Surprisingly, it turns out that there are 
algorithms that can reconstruct user preferences (with high 
probability), using only a little (polylog factor) more questions than 
the minimum possible. 

Short Bio: Boaz Patt-Shamir has been a Professor of Computer Science in Tel Aviv University since 1997, where he directs the laboratory for distributed algorithms. He received his BSc in Mathematics and Computer Science from Tel Aviv University, his MSc from Weizmann Institute, and his PhD from MIT. His interests include distributed network algorithms and algorithms for communication networks. In 2002-2004 he has spent a sabbatical in HP Labs in Cambridge, Massachusetts, where he became interested in recommendation systems. 

Host: Rajmohan Rajaraman