Noisy Sorting Without Resampling

TL;DR

This paper introduces an algorithm for noisy sorting without resampling that efficiently approximates the original order in noisy comparison settings, relevant for ranking applications like sports and search.

Contribution

The paper presents a polynomial-time algorithm with near-optimal sampling complexity for noisy sorting without resampling, and proves the approximate correctness of the obtained order.

Findings

Algorithm runs in n^{O(γ^{-4})} time with high probability.

Sampling complexity is O_γ(n log n).

The output order is close to the original, with bounded displacement.

Abstract

In this paper we study noisy sorting without re-sampling. In this problem there is an unknown order $a_{\pi(1)} < ... < a_{\pi(n)}$ where $\pi$ is a permutation on $n$ elements. The input is the status of $n \choose 2$ queries of the form $q(a_i,x_j)$, where $q(a_i,a_j) = +$ with probability at least $1/2+\ga$ if $\pi(i) > \pi(j)$ for all pairs $i \neq j$, where $\ga > 0$ is a constant and $q(a_i,a_j) = -q(a_j,a_i)$ for all $i$ and $j$. It is assumed that the errors are independent. Given the status of the queries the goal is to find the maximum likelihood order. In other words, the goal is find a permutation $\sigma$ that minimizes the number of pairs $\sigma(i) > \sigma(j)$ where $q(\sigma(i),\sigma(j)) = -$. The problem so defined is the feedback arc set problem on distributions of inputs, each of which is a tournament obtained as a noisy perturbations of a linear order. Note that when $\ga < 1/2$ and $n$ is large, it is impossible to recover the original order $\pi$. It is known that the weighted feedback are set problem on tournaments is NP-hard in general. Here we present an algorithm of running time $n^{O(\gamma^{-4})}$ and sampling complexity $O_{\gamma}(n \log n)$ that with high probability solves the noisy sorting without re-sampling problem. We also show that if $a_{\sigma(1)},a_{\sigma(2)},...,a_{\sigma(n)}$ is an optimal solution of the problem then it is ``close'' to the original order. More formally, with high probability it holds that $\sum_i |\sigma(i) - \pi(i)| = \Theta(n)$ and $\max_i |\sigma(i) - \pi(i)| = \Theta(\log n)$. Our results are of interest in applications to ranking, such as ranking in sports, or ranking of search items based on comparisons by experts.