English

A spectral bound on hypergraph discrepancy

Combinatorics 2020-05-05 v5 Discrete Mathematics Data Structures and Algorithms

Abstract

Let H\mathcal{H} be a tt-regular hypergraph on nn vertices and mm edges. Let MM be the m×nm \times n incidence matrix of H\mathcal{H} and let us denote λ=maxv1,v=1Mv\lambda =\max_{v \perp \overline{1},\|v\| = 1}\|Mv\|. We show that the discrepancy of H\mathcal{H} is O(t+λ)O(\sqrt{t} + \lambda). As a corollary, this gives us that for every tt, the discrepancy of a random tt-regular hypergraph with nn vertices and mnm \geq n edges is almost surely O(t)O(\sqrt{t}) as nn grows. The proof also gives a polynomial time algorithm that takes a hypergraph as input and outputs a coloring with the above guarantee.

Keywords

Cite

@article{arxiv.1907.04117,
  title  = {A spectral bound on hypergraph discrepancy},
  author = {Aditya Potukuchi},
  journal= {arXiv preprint arXiv:1907.04117},
  year   = {2020}
}

Comments

18 pages. arXiv admin note: substantial text overlap with arXiv:1811.01491, several changes to the presentation