Packing Returning Secretaries

TL;DR

This paper investigates online secretary problems with returns in combinatorial packing domains, proposing algorithms with competitive ratios for maximum value packing and analyzing the expected postponements in candidate decision-making.

Contribution

It introduces new algorithms for secretary problems with returns, achieving improved competitive ratios and analyzing postponement bounds in various packing and matroid settings.

Findings

Proposes a 0.5-competitive algorithm for candidates arriving twice.

Achieves at least 0.5721 competitive ratio for bipartite matching with returns.

Shows expected postponements can be reduced to O(r log(n/r)) in matroids.

Abstract

We study online secretary problems with returns in combinatorial packing domains with $n$ candidates that arrive sequentially over time in random order. The goal is to accept a feasible packing of candidates of maximum total value. In the first variant, each candidate arrives exactly twice. All $2n$ arrivals occur in random order. We propose a simple 0.5-competitive algorithm that can be combined with arbitrary approximation algorithms for the packing domain, even when the total value of candidates is a subadditive function. For bipartite matching, we obtain an algorithm with competitive ratio at least $0.5721 - o(1)$ for growing $n$, and an algorithm with ratio at least $0.5459$ for all $n \ge 1$. We extend all algorithms and ratios to $k \ge 2$ arrivals per candidate. In the second variant, there is a pool of undecided candidates. In each round, a random candidate from the pool arrives. Upon arrival a candidate can be either decided (accept/reject) or postponed (returned into the pool). We mainly focus on minimizing the expected number of postponements when computing an optimal solution. An expected number of $\Theta(n \log n)$ is always sufficient. For matroids, we show that the expected number can be reduced to $O(r \log (n/r))$, where $r \le n/2$ is the minimum of the ranks of matroid and dual matroid. For bipartite matching, we show a bound of $O(r \log n)$, where $r$ is the size of the optimum matching. For general packing, we show a lower bound of $\Omega(n \log \log n)$, even when the size of the optimum is $r = \Theta(\log n)$.