Yoram Gat's Home page
Yoram Gat
Xerox Palo Alto Research Center
Theory group, Computer science lab
E-mail address:
gat@parc.xerox.com
Interests:
My research focus is on pattern recognition and classification
problems. I am currently working on problems associated with the
analysis of scanned documents.
In the recent past ('96-'00) I have been a Ph.D. student in the
statistics department of UC Berkeley. My research advisor was
Prof. Peter Bickel.
While in UC Berkeley, I have written two papers regarding the
generalization ability of certain learning algorithms. The first paper
(gzipped postscript)
gives a proof of the good generalization properties of a class of
learning algorithms which includes a version of the "Support
Vector Machine", an algorithm which enjoys some popularity
these days. This paper was accepted to Machine Learning journal.
The second paper (compressed postscript)
gives more precise analysis for some other popular learning algorithms,
such as classification trees and the perceptron.
Both papers were incorporated into my Ph.D. thesis
(compressed postscript).
My M.Sc. Thesis (Tel-Aviv University, Israel) deals
with using partial samples to estimate parameters in influence
diagrams (a.k.a Bayesian networks).
Photos