English

Approximate Selection with Unreliable Comparisons in Optimal Expected Time

Data Structures and Algorithms 2022-05-04 v1

Abstract

Given nn elements, an integer kk and a parameter ε\varepsilon, we study to select an element with rank in (knε,k+nε](k-n\varepsilon,k+n\varepsilon] using unreliable comparisons where the outcome of each comparison is incorrect independently with a constant error probability, and multiple comparisons between the same pair of elements are independent. In this fault model, the fundamental problems of finding the minimum, selecting the kk-th smallest element and sorting have been shown to require Θ(nlog1Q)\Theta\big(n \log \frac{1}{Q}\big), Θ(nlogmin{k,nk}Q)\Theta\big(n\log \frac{\min\{k,n-k\}}{Q}\big) and Θ(nlognQ)\Theta\big(n\log \frac{n}{Q}\big) comparisons, respectively, to achieve success probability 1Q1-Q. Recently, Leucci and Liu proved that the approximate minimum selection problem (k=0k=0) requires expected Θ(ε1log1Q)\Theta(\varepsilon^{-1}\log \frac{1}{Q}) comparisons. We develop a randomized algorithm that performs expected O(knε2log1Q)O(\frac{k}{n}\varepsilon^{-2} \log \frac{1}{Q}) comparisons to achieve success probability at least 1Q1-Q. We also prove that any randomized algorithm with success probability at least 1Q1-Q performs expected Ω(knε2log1Q)\Omega(\frac{k}{n}\varepsilon^{-2}\log \frac{1}{Q}) comparisons. Our results indicate a clear distinction between approximating the minimum and approximating the kk-th smallest element, which holds even for the high probability guarantee, e.g., if k=n2k=\frac{n}{2} and Q=1nQ=\frac{1}{n}, Θ(ε1logn)\Theta(\varepsilon^{-1}\log n) versus Θ(ε2logn)\Theta(\varepsilon^{-2}\log n). Moreover, if ε=nα\varepsilon=n^{-\alpha} for α(0,12)\alpha \in (0,\frac{1}{2}), the asymptotic difference is almost quadratic, i.e., Θ~(nα)\tilde{\Theta}(n^{\alpha}) versus Θ~(n2α)\tilde{\Theta}(n^{2\alpha}).

Keywords

Cite

@article{arxiv.2205.01448,
  title  = {Approximate Selection with Unreliable Comparisons in Optimal Expected Time},
  author = {Shengyu Huang and Chih-Hung Liu and Daniel Rutschman},
  journal= {arXiv preprint arXiv:2205.01448},
  year   = {2022}
}
R2 v1 2026-06-24T11:05:47.630Z