Table of Contents
Fetching ...

Faster $(Δ+ 1)$-Edge Coloring: Breaking the $m \sqrt{n}$ Time Barrier

Sayan Bhattacharya, Din Carmon, Martín Costa, Shay Solomon, Tianyi Zhang

TL;DR

This work addresses the classic problem of computing a $(\Delta+1)$-edge coloring for a simple graph with $n$ vertices and $m$ edges. It introduces a novel canonical-instance framework that isolates a star-centered residual structure, enabling the uncolored edges to be handled via short alternating-path flips and limited fan rotations. By combining a randomized edge-sampling template, an Eulerian-partition-based slack coloring step, and a careful, case-based analysis of Vizing fans, the authors achieve a randomized runtime of $\tilde{O}(mn^{1/3})$, beating the four-decade $m\sqrt{n}$ barrier. The results extend the toolkit for edge coloring, leveraging structure in near-regular subgraphs and a star-aggregation approach that may influence future algorithms for related coloring and scheduling problems. Overall, the paper delivers the first polynomial improvement for this foundational problem and demonstrates a scalable strategy that integrates probabilistic methods, graph partitioning, and Vizing-style color-extension techniques.

Abstract

Vizing's theorem states that any $n$-vertex $m$-edge graph of maximum degree $Δ$ can be {\em edge colored} using at most $Δ+ 1$ different colors [Diskret.~Analiz, '64]. Vizing's original proof is algorithmic and shows that such an edge coloring can be found in $\tilde{O}(mn)$ time. This was subsequently improved to $\tilde O(m\sqrt{n})$, independently by Arjomandi [1982] and by Gabow et al.~[1985]. In this paper we present an algorithm that computes such an edge coloring in $\tilde O(mn^{1/3})$ time, giving the first polynomial improvement for this fundamental problem in over 40 years.

Faster $(Δ+ 1)$-Edge Coloring: Breaking the $m \sqrt{n}$ Time Barrier

TL;DR

This work addresses the classic problem of computing a -edge coloring for a simple graph with vertices and edges. It introduces a novel canonical-instance framework that isolates a star-centered residual structure, enabling the uncolored edges to be handled via short alternating-path flips and limited fan rotations. By combining a randomized edge-sampling template, an Eulerian-partition-based slack coloring step, and a careful, case-based analysis of Vizing fans, the authors achieve a randomized runtime of , beating the four-decade barrier. The results extend the toolkit for edge coloring, leveraging structure in near-regular subgraphs and a star-aggregation approach that may influence future algorithms for related coloring and scheduling problems. Overall, the paper delivers the first polynomial improvement for this foundational problem and demonstrates a scalable strategy that integrates probabilistic methods, graph partitioning, and Vizing-style color-extension techniques.

Abstract

Vizing's theorem states that any -vertex -edge graph of maximum degree can be {\em edge colored} using at most different colors [Diskret.~Analiz, '64]. Vizing's original proof is algorithmic and shows that such an edge coloring can be found in time. This was subsequently improved to , independently by Arjomandi [1982] and by Gabow et al.~[1985]. In this paper we present an algorithm that computes such an edge coloring in time, giving the first polynomial improvement for this fundamental problem in over 40 years.
Paper Structure (28 sections, 7 theorems, 38 equations, 4 figures)

This paper contains 28 sections, 7 theorems, 38 equations, 4 figures.

Key Result

Theorem 1.1

Given a simple undirected graph $G = (V, E)$ on $n$ vertices and $m$ edges with maximum degree $\Delta$, there is a randomized algorithm that computes a $(\Delta+1)$-edge coloring with high probability in runtime $\tilde{O}\left(mn^{\frac{1}{3}}\right)$.

Figures (4)

  • Figure 1: In this picture, A vertex $s$ is directed to $t$ if $\psi(u, t)\in \mathsf{miss}_\psi(s)$. Applying Vizing' procedure for either $(u, v)$ or $(u, v')$ would end up with the same alternating path starting with same $\{x, \psi(u, w_2)\}$-alternating path. We will show that the $\{x, \psi(u, w_1)\}$-alternating path can also be used for color extension.
  • Figure 2: If the neighborhood of $u$ is a star rather than a chain, then we could virtually rotate uncolored edges to $(u, w_0), (u, w_1)$. As $w_0, w_1$ are both missing blue around them, we can pair them together and reuse the argument for the pairing the clients case.
  • Figure 3: If $w$ is the target of many clients in different star subgraphs, we can show that all these star centers have the same type of alternating paths which begin with blue edges.
  • Figure 8: This example shows a tree component $T$ where $D\setminus F$ are red vertices, $F$ are orange vertices, and branching vertices are blue ones. So by definition, $\mathsf{br}_\psi(T) = 4 = 1+|D\setminus F|$.

Theorems & Definitions (41)

  • Theorem 1.1
  • Lemma 2.1
  • Lemma 3.1
  • proof
  • Lemma 3.2: Eulerian partition arjomandi1982efficientgabow1985algorithmssinnamon2019fast
  • Lemma 3.3: elkin2024deterministic
  • proof
  • Theorem 4.1
  • Claim 4.1
  • proof
  • ...and 31 more