English

Centralized, Parallel, and Distributed Multi-Source Shortest Paths via Hopsets and Rectangular Matrix Multiplication

Data Structures and Algorithms 2021-02-15 v2

Abstract

Consider an undirected weighted graph G=(V,E,w)G = (V,E,w). We study the problem of computing (1+ϵ)(1+\epsilon)-approximate shortest paths for S×VS \times V, for a subset SVS \subseteq V of S=nr|S| = n^r sources, for some 0<r10 < r \le 1. We devise a significantly improved algorithm for this problem in the entire range of parameter rr, in both the classical centralized and the parallel (PRAM) models of computation, and in a wide range of rr in the distributed (Congested Clique) model. Specifically, our centralized algorithm for this problem requires time O~(Eno(1)+nω(r))\tilde{O}(|E| \cdot n^{o(1)} + n^{\omega(r)}), where nω(r)n^{\omega(r)} is the time required to multiply an nr×nn^r \times n matrix by an n×nn \times n one. Our PRAM algorithm has polylogarithmic time (logn)O(1/ρ)(\log n)^{O(1/\rho)}, and its work complexity is O~(Enρ+nω(r))\tilde{O}(|E| \cdot n^\rho + n^{\omega(r)}), for any arbitrarily small constant ρ>0\rho >0. In particular, for r0.313r \le 0.313\ldots, our centralized algorithm computes S×VS \times V (1+ϵ)(1+\epsilon)-approximate shortest paths in n2+o(1)n^{2 + o(1)} time. Our PRAM polylogarithmic-time algorithm has work complexity O(Enρ+n2+o(1))O(|E| \cdot n^\rho + n^{2+o(1)}), for any arbitrarily small constant ρ>0\rho >0. Previously existing solutions either require centralized time/parallel work of O(ES)O(|E| \cdot |S|) or provide much weaker approximation guarantees. In the Congested Clique model, our algorithm solves the problem in polylogarithmic time for S=nr|S| = n^r sources, for r0.655r \le 0.655, while previous state-of-the-art algorithms did so only for r1/2r \le 1/2. Moreover, it improves previous bounds for all r>1/2r > 1/2. For unweighted graphs, the running time is improved further to poly(loglogn)poly(\log\log n).

Keywords

Cite

@article{arxiv.2004.07572,
  title  = {Centralized, Parallel, and Distributed Multi-Source Shortest Paths via Hopsets and Rectangular Matrix Multiplication},
  author = {Michael Elkin and Ofer Neiman},
  journal= {arXiv preprint arXiv:2004.07572},
  year   = {2021}
}
R2 v1 2026-06-23T14:53:32.398Z