English

On the subgraph query problem

Combinatorics 2021-01-27 v2

Abstract

Given a fixed graph HH, a real number p(0,1)p\in(0,1), and an infinite Erd\H{o}s-R\'enyi graph GG(,p)G\sim G(\infty,p), how many adjacency queries do we have to make to find a copy of HH inside GG with probability 1/21/2? Determining this number f(H,p)f(H,p) is a variant of the {\it subgraph query problem} introduced by Ferber, Krivelevich, Sudakov, and Vieira. For every graph HH, we improve the trivial upper bound of f(H,p)=O(pd)f(H,p) = O(p^{-d}), where dd is the degeneracy of HH, by exhibiting an algorithm that finds a copy of HH in time o(pd)o(p^{-d}) as pp goes to 00. Furthermore, we prove that there are 22-degenerate graphs which require p2+o(1)p^{-2+o(1)} queries, showing for the first time that there exist graphs HH for which f(H,p)f(H,p) does not grow like a constant power of p1p^{-1} as pp goes to 00. Finally, we answer a question of Feige, Gamarnik, Neeman, R\'acz, and Tetali by showing that for any δ<2\delta < 2, there exists α<2\alpha < 2 such that one cannot find a clique of order αlog2n\alpha \log_2 n in G(n,1/2)G(n,1/2) in nδn^\delta queries.

Keywords

Cite

@article{arxiv.1911.04413,
  title  = {On the subgraph query problem},
  author = {Ryan Alweiss and Chady Ben Hamida and Xiaoyu He and Alexander Moreira},
  journal= {arXiv preprint arXiv:1911.04413},
  year   = {2021}
}

Comments

modified slightly after reviewer comments