English

Approximate Minimum Selection with Unreliable Comparisons in Optimal Expected Time

Data Structures and Algorithms 2020-07-08 v3

Abstract

We consider the \emph{approximate minimum selection} problem in presence of \emph{independent random comparison faults}. This problem asks to select one of the smallest kk elements in a linearly-ordered collection of nn elements by only performing \emph{unreliable} pairwise comparisons: whenever two elements are compared, there is a constant probability that the wrong answer is returned. We design a randomized algorithm that solves this problem with probability 1q[12,1)1-q \in [ \frac{1}{2}, 1) and for the whole range of values of kk using O(nklog1q)O( \frac{n}{k} \log \frac{1}{q} ) expected time. Then, we prove that the expected running time of any algorithm that succeeds w.h.p. must be Ω(nklog1q)\Omega(\frac{n}{k}\log \frac{1}{q}), thus implying that our algorithm is asymptotically optimal, in expectation. These results are quite surprising in the sense that for kk between Ω(log1q)\Omega(\log \frac{1}{q}) and cnc \cdot n, for any constant c<1c<1, the expected running time must still be Ω(nklog1q)\Omega(\frac{n}{k}\log \frac{1}{q}) even in absence of comparison faults. Informally speaking, we show how to deal with comparison errors without any substantial complexity penalty w.r.t.\ the fault-free case. Moreover, we prove that as soon as k=O(nloglog1q)k = O( \frac{n}{\log\log \frac{1}{q}}), it is possible to achieve the optimal \emph{worst-case} running time of Θ(nklog1q)\Theta(\frac{n}{k}\log \frac{1}{q}).

Keywords

Cite

@article{arxiv.1805.02033,
  title  = {Approximate Minimum Selection with Unreliable Comparisons in Optimal Expected Time},
  author = {Stefano Leucci and Chih-Hung Liu},
  journal= {arXiv preprint arXiv:1805.02033},
  year   = {2020}
}

Comments

13 pages

R2 v1 2026-06-23T01:45:53.947Z