AVERAGE-CASE AND SMOOTHED ANALYSIS OF GRAPH ISOMORPHISM

Julia Gaudio, Miklós Z. Rácz, Anirudh Sridhar

Research output: Contribution to journalArticlepeer-review

Abstract

We propose a simple and efficient local algorithm for graph isomorphism which succeeds for a large class of sparse graphs. This algorithm produces a low-depth canonical labeling, which is a labeling of the vertices of the graph that identifies its isomorphism class using vertices’ local neighborhoods. Prior work by Czajka and Pandurangan showed that in the Erdős–Rényi model G(n, pn), the degree profile of a vertex (i.e., the sorted list of the degrees of its neighbors) gives a canonical labeling with high probability when npn = ω(log4(n)/log log n) (and pn ≤ 1/2); subsequently, Mossel and Ross showed that the same holds when npn = ω(log2(n)). We first show that their analysis essentially cannot be improved: we prove that when npn = o(log2(n)/(log log n)3), with high probability there exist distinct vertices with isomorphic 2-neighborhoods. Our first main result is a positive counterpart to this, showing that 3-neighborhoods give a canonical labeling when npn ≥ (1 + δ)log n (and pn ≤ 1/2); this improves a recent result of Ding, Ma, Wu and Xu, completing the picture above the connectivity threshold. Our second main result is a smoothed analysis of graph isomorphism, showing that for a large class of deterministic graphs, a small random perturbation of the edge set yields a graph which admits a canonical labeling from 3-neighborhoods, with high probability. While the worst-case complexity of graph isomorphism is still unknown, this shows that graph isomorphism has polynomial smoothed complexity.

Original languageEnglish (US)
Pages (from-to)1373-1406
Number of pages34
JournalAnnals of Applied Probability
Volume35
Issue number2
DOIs
StatePublished - Apr 2025

Funding

We thank Lutz Warnke for observing that the method of typical bounded differences simplifies the proof of Lemma 4.1 (and thus also the proof of Lemma 6.1). We thank Aravindan Vijayaraghavan for helpful discussions, as well as the anonymous reviewers of both an earlier version of this work and the current version, who all gave many helpful comments that have improved the paper. A.S. acknowledges support from the U.S. National Science Foundation under Grants CCF-1908308, ECCS-2039716 and CCF-1918421, the Office of Naval Research under Grant ONR-N00014-20-1-2826, as well as a grant from the c3.ai Digital Transformation Institute.

Keywords

  • canonical labeling
  • graph isomorphism
  • graph shotgun assembly
  • Random graphs
  • randomly perturbed graphs
  • smoothed analysis

ASJC Scopus subject areas

  • Statistics and Probability
  • Statistics, Probability and Uncertainty

Fingerprint

Dive into the research topics of 'AVERAGE-CASE AND SMOOTHED ANALYSIS OF GRAPH ISOMORPHISM'. Together they form a unique fingerprint.

Cite this