Dear all,
I would like to announce the following talk that will be held on Thursday January 23rd 2020, at 4:30 PM, in Aula D’Antoni of the Department of Mathematics, University of Rome Tor Vergata.
Speaker: Ernesto De Vito
Title: Multi-Scale Vector Quantization with Reconstruction Trees
Abstract: The aim of vector quantization is a parsimonious reconstruction of unsupervised data as provided by K-means and geometric multi-resolution analysis, to name a few. In the talk I will present a multi-scale algorithm for vector quantization and study its statistical properties. The proposed approach is inspired by decision trees and is based a family of given partitions, to quickly explore the data in a coarse to fine-multi-scale-fashion. The main technical contribution is an analysis of the expected distortion achieved by the proposed algorithm. Both asymptotic and finite sample results are derived. The presentation is based on a joint work with E. Cecini and L. Rosasco.