[Ml-stat-talks] PACM Colloquium Nov 14, 2011 - Vladimir Rokhlin

Amit Singer amits at math.Princeton.EDU
Wed Nov 9 21:45:50 EST 2011



I would like to bring to your attention the upcoming PACM Colloquium by
Vladimir Rokhlin (see below).

For those of you who don't know Vladimir, he is a member of both the
National Academy of Sciences and the National Academy of Engineering, and
perhaps mostly known as the co-inventor of the fast multiple method.

While not clear from his abstract and title, Vladimir will discuss at least
two algorithms of great importance for almost anyone working with large data
sets: the partial SVD (i.e., finding low-rank approximation) of large-scale
matrices and approximate nearest neighbors search.





From: Valerie Marino [mailto:vmarino at math.princeton.edu] 
Sent: Tuesday, November 08, 2011 9:53 AM
To: pacm-colloquium at math.princeton.edu
Cc: vmarino at math.princeton.edu
Subject: PACM Colloquium Nov, 14, 2011


DATE: Monday, November 14, 2011 

PLACE: 214 Fine Hall 

TIME: 4:30 pm 

SPEAKER: Vladimir Rokhlin, Yale University

TITLE: Accurate Randomized Algorithms of Numerical Analysis

Randomized algorithms are ubiqutous in computer science and computer
engineering. Many problems that are intractable when viewed deterministi-
cally can be effectively solved with probabilistic techniques. Perhaps the
most important aspect of most randomized procedures in current use is the
fact that they produce the correct result with (practically speaking) 100%
reliabil- ity, and with (essentially) machine precision. Historically,
randomized techniques have been less popular in numerical analysis. Most of
them trade accuracy for speed, and in many numerical environments one does
not want to add yet another source of inaccuracy to the calculation that is
already sufficiently inaccurate. One could say that in numerical analysis
probabilistic methods are an approach of last resort. I will discuss several
probabilistic algorithms of numerical linear algebra that are never less
accurate than their deterministic counterparts, and in fact tend to produce
better accuracy. In many situations, the new schemes have lower CPU time
requirements than existing methods, both asymptotically and in terms of
actual timings. I will illustrate the approach with several numerical
examples, and discuss possible extensions. 

Coffee & Tea will be served at 3:55 p.m. in 217A Fine Hall

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/ml-stat-talks/attachments/20111109/b5fbb075/attachment.html>

More information about the Ml-stat-talks mailing list