Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

On Approximability of Steiner Tree in $\ell_p$-metrics

Henry Fleischmann, Surya Teja Gavva, Karthik C. S

TL;DR

This work establishes APX-hardness of Discrete Steiner Tree (DST) across all $\ell_p$-metrics and APX-hardness of Continuous Steiner Tree (CST) in the $\ell_{\infty}$-metric, resolving fundamental gaps in the hardness landscape for high-dimensional geometric Steiner problems. It introduces a unified framework that reduces CST to DST in $\ell_p$-spaces and builds a spectrum of gap-preserving reductions from Set Packing (SP3) to DST in Euclidean and general $\ell_p$-metrics, yielding explicit inapproximability factors dependent on $p$. The authors also demonstrate APX-hardness for DST in string metrics via isometric embeddings (Hamming to Ulam and edit distance) and show that hardness persists under dimensionality reduction to $O(\log n)$ dimensions, assuming appropriate complexity hypotheses. Additionally, a CST-to-DST reduction grounded in recent structural results on near-optimal Steiner trees links the two problems and yields a practical route to translate DST hardness into CST limits, with implications for approximation algorithms (e.g., a $1.39$-approximation for high-dimensional CST given DST solvers). Overall, the paper significantly strengthens the hardness landscape for Steiner tree variants and provides versatile reductions and embeddings that connect discrete, continuous, and geometric formulations across metrics.

Abstract

In the Continuous Steiner Tree problem (CST), we are given as input a set of points (called terminals) in a metric space and ask for the minimum-cost tree connecting them. Additional points (called Steiner points) from the metric space can be introduced as nodes in the solution. In the Discrete Steiner Tree problem (DST), we are given in addition to the terminals, a set of facilities, and any solution tree connecting the terminals can only contain the Steiner points from this set of facilities. Trevisan [SICOMP'00] showed that CST and DST are APX-hard when the input lies in the $\ell_1$-metric (and Hamming metric). Chlebík and Chlebíková [TCS'08] showed that DST is NP-hard to approximate to factor of $96/95\approx 1.01$ in the graph metric (and consequently $\ell_\infty$-metric). Prior to this work, it was unclear if CST and DST are APX-hard in essentially every other popular metric. In this work, we prove that DST is APX-hard in every $\ell_p$-metric. We also prove that CST is APX-hard in the $\ell_{\infty}$-metric. Finally, we relate CST and DST, showing a general reduction from CST to DST in $\ell_p$-metrics.

On Approximability of Steiner Tree in $\ell_p$-metrics

TL;DR

This work establishes APX-hardness of Discrete Steiner Tree (DST) across all -metrics and APX-hardness of Continuous Steiner Tree (CST) in the -metric, resolving fundamental gaps in the hardness landscape for high-dimensional geometric Steiner problems. It introduces a unified framework that reduces CST to DST in -spaces and builds a spectrum of gap-preserving reductions from Set Packing (SP3) to DST in Euclidean and general -metrics, yielding explicit inapproximability factors dependent on . The authors also demonstrate APX-hardness for DST in string metrics via isometric embeddings (Hamming to Ulam and edit distance) and show that hardness persists under dimensionality reduction to dimensions, assuming appropriate complexity hypotheses. Additionally, a CST-to-DST reduction grounded in recent structural results on near-optimal Steiner trees links the two problems and yields a practical route to translate DST hardness into CST limits, with implications for approximation algorithms (e.g., a -approximation for high-dimensional CST given DST solvers). Overall, the paper significantly strengthens the hardness landscape for Steiner tree variants and provides versatile reductions and embeddings that connect discrete, continuous, and geometric formulations across metrics.

Abstract

In the Continuous Steiner Tree problem (CST), we are given as input a set of points (called terminals) in a metric space and ask for the minimum-cost tree connecting them. Additional points (called Steiner points) from the metric space can be introduced as nodes in the solution. In the Discrete Steiner Tree problem (DST), we are given in addition to the terminals, a set of facilities, and any solution tree connecting the terminals can only contain the Steiner points from this set of facilities. Trevisan [SICOMP'00] showed that CST and DST are APX-hard when the input lies in the -metric (and Hamming metric). Chlebík and Chlebíková [TCS'08] showed that DST is NP-hard to approximate to factor of in the graph metric (and consequently -metric). Prior to this work, it was unclear if CST and DST are APX-hard in essentially every other popular metric. In this work, we prove that DST is APX-hard in every -metric. We also prove that CST is APX-hard in the -metric. Finally, we relate CST and DST, showing a general reduction from CST to DST in -metrics.
Paper Structure (41 sections, 50 theorems, 22 equations, 5 figures)

This paper contains 41 sections, 50 theorems, 22 equations, 5 figures.

Table of Contents

  1. Introduction
  2. General Metrics.
  3. Geometric Metrics.
  4. Our results
  5. String Metrics.
  6. Dimensionality Reduction.
  7. Our Techniques
  8. Proof Overview of Theorem \ref{['thm:introellinf']}: Hardness of Approximation of CST in linf-metric
  9. Prior Work on CST and Technical Difficulties.
  10. Reduction from Graph Coloring to CST in the l_inf-metric.
  11. Proof Overview of Theorems \ref{['thm:introapx']} and \ref{['thm:intromain']}: Hardness of Approximation of DST in lp-metrics
  12. Known Reductions from Set Cover and Vertex Cover.
  13. Reduction from Set Packing to Euclidean DST.
  14. Reduction from Set Packing to DST in lp-metric Spaces.
  15. Proof Overview of Theorem \ref{['thm:csttodstintro']}: Reduction from CST to DST
  16. ...and 26 more sections

Key Result

Theorem 1.1

There is an efficiently computable function mapping a graph $G$ of order $n$ to an instance of $\mathsf{CST}$ in the $\ell_{\infty}$-metric such that the optimal cost of the Steiner tree is $(n + \chi(G))/2$, where $\chi(G)$ is the chromatic number of $G$. Consequently, $\mathsf{CST}$ in the $\ell_{

Figures (5)

  • Figure 4: Color-coded range of $\ell_p$ hardness of approximation dependency on $\varepsilon$ and $\delta$.
  • Figure 5: Instance of $\mathsf{DST}$ in general metrics from Karp's reduction from Set Cover
  • Figure 6: Steiner tree of $P$ constructed from an $a$-coloring of $G$
  • Figure 7: A general $(\alpha_{\text{X}}, \alpha_{\text{P}}, \beta_{\text{in}}, \beta_{\text{out}}, \gamma_0, \gamma_1, \gamma_2, \tau)$-set system space with edges between different groups of nodes color-coded
  • Figure 8: Steiner tree construction in the completeness case of Theorem \ref{['thm: graph hardness']}.

Theorems & Definitions (59)

  • Theorem 1.1
  • Theorem 1.2: Implied by Theorem \ref{['thm: hardness of lp dst']}
  • Theorem 1.3
  • Theorem 1.4: Implied by Theorem \ref{['thm: reduction from cst to dst']}
  • Theorem 1.5
  • Theorem 1.6
  • remark 2.1
  • Theorem 2.2: Theorem 4.4, Petrank_1994
  • Theorem 3.1: Cerioli_Faria_Ferreira_Martinhon_Protti_Reed_2008, Theorem 7
  • Lemma 3.2
  • ...and 49 more