PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

On codecell convexity of optimal multi-resolution scalar quantizers for continuous sources
András Antos
IEEE Transactions on Information Theory 2009. ISSN 0018-9448

Abstract

It has been shown by earlier results that codecell convexity may preclude optimality for multiple-description and for multi-resolution scalar quantizers (MRSQ). Given that the later result is proved showing a discrete counter-example, it could be suspected that for the class of sources with uniformly bounded continuous densities, there is always a convex optimal MRSQ. Here it is shown that this is not so, proving that convexity of fixed-rate MRSQs may preclude optimality also for these subclasses.

EPrint Type:Article
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Learning/Statistics & Optimisation
Theory & Algorithms
ID Code:6025
Deposited By:András Antos
Deposited On:08 March 2010