On Convergence Rate of the Generalized Diversity Subsampling Method

TL;DR

This paper analyzes the convergence rate of the generalized Diversity Subsampling (g-DS) method, focusing on how density estimation errors affect its asymptotic performance when selecting samples from large datasets.

Contribution

It provides a theoretical analysis of the pointwise convergence rate of g-DS considering density estimation bias and variance, extending previous asymptotic results.

Findings

Convergence rate depends on density estimation bias and variance.

Density estimation errors influence the asymptotic performance of g-DS.

Theoretical bounds on the convergence rate are established.

Abstract

arXiv:2206.10812v1 [stat.ME] proposes a useful algorithm, named generalized Diversity Subsampling (g-DS) algorithm, to select a subsample following some target probability distribution from a finite data set and demonstrates its effectiveness numerically. While the asymptotic performances of g-DS when the true data distribution is known was discussed in arXiv:2206.10812v1 [stat.ME], it remains an interesting question how the estimation errors in the density estimation step, which is an unavoidable step to use g-DS in real-world data sets, influences its asymptotic performance. In this paper, we study the pointwise convergence rate of probability density function (p.d.f) the g-DS subsample to the target p.d.f value, as the data set size approaches infinity, under consideration of the pointwise bias and variance of the estimated data p.d.f.