STATISTICS COLLOQUIUM

THURSDAY, April 23rd, 2009 at 4.10pm, MSB 1147 (Colloquium Room)
Refreshments: 3.30pm, MSB 4110 (Statistics Lounge)

Speaker: Bruno Pelletier (Univ. Montpellier II, France, visiting UC San Diego)

Title: Clustering with level sets

Abstract: The objective of clustering, or unsupervised classification, is to partition a set of observations into different groups, or clusters, based on their similarities. Following Hartigan, a cluster is defined as a connected component of an upper level set of the underlying density. In this talk, we introduce a spectral clustering algorithm on estimated level sets, and we establish its strong consistency. We also discuss the estimation of the number of connected components of density level sets.