| DATE: | Thursday, May 10, 2001 |
| TIME: | 4:30-5:30 PM |
| PLACE: | LN 2205 |
| SPEAKER: | Evarist Gine |
| TITLE: | Applications of empirical processes in density estimation |
We will discuss how to obtain the a.s. exact asymptotic behavior of the maximum discrepancy over $R^d$ between a kernel density estimator and its expectation, using modern empirical process theory. The main tool for this is a now famous exponential inequality of Talagrand which is proving to be useful in many areas. If time allows, we will also look at the $L_1$ norm of the discrepancy.