[Paper Review] Vertex identifying codes for the n-dimensional lattice

Brendon Stanton|arXiv (Cornell University)|Aug 28, 2010

Digital Image Processing Techniques5 references4 citations

TL;DR

This paper establishes the asymptotic density of r-identifying codes on the n-dimensional integer lattice Lₙ, proving that for fixed n, the minimum density is Θ(1/r^{n-1}) as r → ∞. It constructs a sparse 1-identifying code for the 4-dimensional lattice with density 2/9 and generalizes bounds using dominating sets and geometric covering arguments.

ABSTRACT

An $r$-identifying code on a graph $G$ is a set $C\subset V(G)$ such that for every vertex in $V(G)$, the intersection of the radius-$r$ closed neighborhood with $C$ is nonempty and different. Here, we provide an overview on codes for the $n$-dimensional lattice, discussing the case of 1-identifying codes, constructing a sparse code for the 4-dimensional lattice as well as showing that for fixed $n$, the minimum density of an $r$-identifying code is $Θ(1/r^{n-1})$.

Motivation & Objective

• To determine the asymptotic minimum density of r-identifying codes on the n-dimensional integer lattice Lₙ for fixed n.
• To construct explicit sparse 1-identifying codes for small dimensions, particularly for the 4-dimensional lattice.
• To generalize bounds using dominating sets in hypercubes and apply results from Kabatyanskiĭ and Panchenko for asymptotic analysis.
• To establish tight upper and lower bounds on the density of r-identifying codes, showing Θ(1/r^{n-1}) scaling.

Proposed method

• Uses a geometric embedding of the 2D king grid code into 4D lattice via linearly independent vectors (1,1,1,0) and (1,−1,0,1), preserving identifying properties.
• Applies the concept of dominating sets in hypercubes to bound the 1-identifying code density for Lₙ, generalizing Hamming code constructions.
• Employs a covering argument using balls of radius r in Lₙ, partitioning space into cubes of side length k to bound the number of codewords per unit volume.
• Uses unimodal distance functions and codeword spacing to reconstruct vertex coordinates from identifying sets, enabling unique identification.
• Applies the pigeonhole principle and geometric constraints to derive lower bounds on code density via volume and covering arguments.
• Uses asymptotic analysis with r₀ = floor(r(n+2)/2) to ensure sufficient codeword spacing and coordinate reconstruction accuracy.

Experimental results

Research questions

• RQ1 What is the minimum possible density of an r-identifying code on the n-dimensional lattice Lₙ for fixed n?
• RQ2 Can a sparse 1-identifying code be constructed for the 4-dimensional lattice, and what is its density?
• RQ3 How does the density of r-identifying codes scale with r for fixed n?
• RQ4 What is the relationship between dominating sets in hypercubes and the density of 1-identifying codes in Lₙ?
• RQ5 How tight are the upper and lower bounds on code density, and what is their asymptotic ratio?

Key findings

• For fixed n, the minimum density of an r-identifying code on Lₙ is Θ(1/r^{n-1}) as r → ∞, establishing the exact asymptotic scaling.
• A 1-identifying code for the 4-dimensional lattice L₄ with density 2/9 is explicitly constructed by lifting a known 2D king grid code.
• The 1-identifying code density for L₄ satisfies 1/5 ≤ D(L₄,1) ≤ 2/9, with the upper bound proven via the 4D embedding construction.
• For large n, the 1-identifying code density satisfies 1/(n+1) ≤ D(Lₙ,1) ≤ (1 + b ln ln n / ln n)/ (n+1) for some constant b.
• The ratio of the upper to lower bound for code density is asymptotically bounded by approximately 2e^{n+1}/(√(2πn)⌈log₂(2n+1)⌉), indicating a gap growing as eⁿ.
• The method enables unique reconstruction of vertex coordinates from identifying sets using codeword spacing and distance minimization, ensuring distinguishability.