Asymptotics of the geometric mean error for in-homogeneous self-similar measures

TL;DR

This paper investigates the asymptotic behavior of geometric mean errors in quantization for in-homogeneous self-similar measures, establishing the quantization dimension and convergence rates under specific self-similarity and open set conditions.

Contribution

It proves the existence of the quantization dimension for in-homogeneous self-similar measures and determines the precise convergence order of geometric mean errors.

Findings

Quantization dimension equals that of the associated self-similar measure.

Geometric mean errors decay at a rate proportional to n^{-1/d_0}.

Results hold under in-homogeneous open set conditions.

Abstract

Let $(f_i)_{i=1}^N$ be a family of contractive similitudes on $\mathbb{R}^q$ satisfying the open set condition. Let $(p_i)_{i=0}^N$ be a probability vector with $p_i>0$ for all $i=0,1,\ldots,N$. We study the asymptotic geometric mean errors $e_{n,0}(\mu),n\geq 1$, in the quantization for the in-homogeneous self-similar measure $\mu$ associated with the condensation system $((f_i)_{i=1}^N,(p_i)_{i=0}^N,\nu)$. We focus on the following two independent cases: (I) $\nu$ is a self-similar measure on $\mathbb{R}^q$ associated with $(f_i)_{i=1}^N$; (II) $\nu$ is a self-similar measure associated with another family of contractive similitudes $(g_i)_{i=1}^M$ on $\mathbb{R}^q$ satisfying the open set condition and $((f_i)_{i=1}^N,(p_i)_{i=0}^N,\nu)$ satisfies a version of in-homogeneous open set condition. We show that, in both cases, the quantization dimension $D_0(\mu)$ of $\mu$ of order zero exists and agrees with that of $\nu$, which is independent of the probability vector $(p_i)_{i=0}^N$. We determine the convergence order of $(e_{n,0}(\mu))_{n=1}^\infty$; namely, for $D_0(\mu)=:d_0$, there exists a constant $D>0$, such that \[ D^{-1}n^{-\frac{1}{d_0}}\leq e_{n,0}(\mu)\leq D n^{-\frac{1}{d_0}}, n\geq 1. \]