B1138
Title: Clustering via local depth functions
Authors: Giacomo Francisci - George Mason University (United States) [presenting]
Claudio Agostinelli - University of Trento (Italy)
Alicia Nieto-Reyes - Universidad de Cantabria (Spain)
Anand Vidyashankar - George Mason University (United States)
Abstract: Depth functions are used to identify a center for multivariate distributions. Local depth functions (LDFs) involve an additional tuning parameter and are used to identify local features of the distribution such as peaks and valleys. When the tuning parameter converges to zero, appropriately rescaled LDFs converge to the underlying density and can be used in a modal clustering algorithm. We show that, as the sample size tends to infinity, empirical clusters converge to the corresponding population clusters.
