Print

Print


[Apologies for multiple postings]

**************************************************
*
*
*    GRAND Seminar
*
*    http://cs.gmu.edu/~robotics/Main/GrandSeminar
*
*
**************************************************


*Title*

Efficient solutions for large scale learning:
applications in speaker recognition and geostatistics

*Time/Venue*

Nov/09/2010, Tuesday noon-1:00pm. ENGR 4201

*Speaker*

Balaji Vasan
http://www.umiacs.umd.edu/users/balajiv/home.htm

PhD candidate
Department of Computer Science
University of Maryland

*Host*

Prof. Huzefa Rangwala

*Abstract*

With the ease of data collection, the amount of data available for
learning has increased by several folds. This requires any learning
technique being used to scale well to large data with many
attributes/features. This talk will focus on new machine learning
solutions that address the scalability. The applications considered
include weather data modeling, speaker recognition and computer
vision.

The talk will be divided into two parts; in the first half, the focus
will be on Gaussian process regression (GPR). Gaussian process
regression is a non-parametric learning technique that has been proven
to be robust, but is hindered by its high computational cost.
Acceleration of GPR on graphical processors and iterative Krylov
solvers will be presented and the framework would be extended to an
efficient geostatistical kriging for weather data interpolation. The
second half of the talk will focus on a new partial least squares
(PLS) regression framework for speaker recognition. PLS has already
been applied in several computer vision problems, its extension to
speaker recognition in a “supervector” space will be discussed. The
new framework has been accelerated on graphical processors, as well.


-- 
*Jyh-Ming Lien*
Assistant Professor, George Mason University
+1-703-993-9546
http://cs.gmu.edu/~jmlien