July 26, 2001

BACK


SHAPE - Seminar

Statistical Hypothesis Testing with Possibly Mis-specified and Non-Nested Statistical Pattern Recognition Models

Professor Richard M. Golden

University of Texas at Dallas

Monday July 30, 2001, 11am

Barus-Holley, Room 190


Abstract

Asymptotic statistical theory is used to develop statistical tests for statistical pattern recognition algorithms designed to process stationary time-series data. Particular difficulties in applying such techniques to artificial neural networks are discussed and some approaches for dealing with these difficulties are briefly reviewed. In addition, a new Statistical Test for comparing two non-nested probability models which may be mis-specified is also introduced.


BACK

Last Updated: July 26, 2001