LAGO: Efficient RBFnets for Rare Target Detection
Mu Zhu (University of Waterloo, Ontario, Canada)
MSI Computational Mathematics (formerly AdvCom) Seminar SeriesDATE: 2006-10-30
TIME: 10:30:00 - 11:30:00
LOCATION: John Dedman Seminar Room G35
CONTACT: JavaScript must be enabled to display this email address.
ABSTRACT:
In this talk, I will focus on a general class of statistical problems where the underlying objective is to detect items belonging to a rare class from a very large database. I will introduce an efficient computational algorithm called LAGO. In theory, I will show that LAGO can be justified as an adaptive-bandwidth kernel density estimate of the rare class density that is then adjusted locally by a factor which approximates the background class density to the first order. I will also argue that LAGO is a highly efficient way to construct a radial basis function network (RBFnet) for the rare target detection problem.
BIO:
http://www.math.uwaterloo.ca/~m3zhu/cover.html
