Seminar: Slepians Lemma Light

SpeakerAndreas Maurer
DateWednesday, 28 Nov 2012
Time12:00 - 13:00
LocationMPEB 1.04
Event seriesComputer Science Seminars
Description

Slepian's Lemma is a nontrivial result on Gaussian processes which can be applied to prove uniform convergence bounds for neural networks, transfer learning and various methods of unsupervised learning. It turns out that for these applications one just needs a special consequence of Slepian's inequality, which has a much simpler and direct proof. The talk introduces the problem in the case of K-means clustering, states and proves the corresponding result and gives some applications.

iCalendar csml_id_127.ics