Data Intelligence, Business Analytics
Date: Monday March 22, 2010; 6:30 pm
Speakers: Andrea Montanari, Stanford Professor in Electrical Engineering and Statistics
Title: “Large Matrices beyond Singular Value Decomposition”
A number of data sets are naturally described in matrix form. Examples range from micro-arrays to collaborative filtering data. In many
of these examples, singular value decomposition (SVD) provides an efficient way
to construct a low-rank approximation thus achieving a large dimensionality
reduction. SVD is also an important tool in the design of approximate
linear algebra algorithms for massive data sets. It is a recent discovery
that –for ‘generic’ matrices — SVD is sub-optimal, and can be significantly
improved upon. There has been considerable progress on this topic over
the last year, partly spurred by interest in the Netflix challenge. I
will overview this progress.
He was co-awarded the ACM SIGMETRICS best paper award in 2008. He received the CNRS bronze medal for theoretical physics in 2006 and the National Science Foundation CAREER award in 2008.