Tuesday April 10, 2018 3:00 p.m. Yost 306

Speaker: Manfred Danker, Pennsylvania State University

Title: Improving Statistical Decision Procedures

Abstract: I shall discuss a smoothing method for distributions without changing expectations while forcing convergence to some distribution when applied to i.i.d. samples with finite 1 +δ moments for some δ > 0. Among others, this applies to sampling when the variance is huge. As an example I shall discuss the problem of estimating the coupling strength in fully coupled neural networks.

Partners: Dr. Elisabeth Werner, Dr. Mark Meckes, Dr. Elizabeth Meckes