[Paper Review] Linear time Constructions of some $d$-Restriction Problems

Nader H. Bshouty|arXiv (Cornell University)|Jun 9, 2014

Complexity and Algorithms in Graphs21 references4 citations

TL;DR

This paper presents the first linear-time, globally explicit constructions for three $d$-restriction problems—perfect hash families, cover-free families, and separating hash families—using algebraic techniques based on finite fields and hitting sets. The key contribution is achieving optimal or near-optimal sizes with deterministic, linear-time construction, improving upon prior polynomial-time or less efficient methods.

ABSTRACT

We give new linear time globally explicit constructions for perfect hash families, cover-free families and separating hash functions.

Motivation & Objective

• To develop efficient, deterministic constructions for $d$-restriction problems with minimal size and linear-time complexity.
• To address the open problem of constructing almost optimal perfect hash families for $q = o(d^2)$, particularly when $q$ is small relative to $d$.
• To unify and improve existing constructions of cover-free and separating hash families using algebraic geometry and hitting set theory.
• To provide globally explicit constructions (running in linear time) that are smaller than known prior constructions.

Proposed method

• Leverages finite fields and polynomial constructions over $\mathbb{F}_q$ to generate hitting sets for hypergraphs of $d$-tuples.
• Uses the product of differences $\prod_{1\leq k<j\leq d}(x_{i_k} - x_{i_j})$ to define separating polynomials that ensure injectivity or coverage.
• Applies hitting set theory to reduce the size of hash families by selecting minimal sets of evaluations that satisfy all required constraints.
• Employs Reed-Solomon-like codes and algebraic function fields to construct explicit families with controlled size and structure.
• Derives bounds using union bound and probabilistic methods to establish near-optimality of the resulting constructions.
• Introduces a framework to transform hitting sets for hypergraphs into compact, globally explicit constructions for all three $d$-restriction problems.

Experimental results

Research questions

• RQ1 Can we construct $d$-restriction problems in linear time with sizes approaching the theoretical lower bounds?
• RQ2 What is the minimal size of a perfect hash family for $q = \Omega(d^2)$, and can it be constructed deterministically in linear time?
• RQ3 How can we achieve almost optimal constructions for $q = o(d^2)$, especially when prior methods are inefficient or suboptimal?
• RQ4 Can global explicit constructions for cover-free and separating hash families be made both efficient and near-optimal in size?
• RQ5 What role do algebraic structures like finite fields and polynomial ideals play in enabling linear-time construction of $d$-restriction families?

Key findings

• For $q > 4(d(d-1)/2 + 1)$, a linear-time construction of an $(n,q,d)$-perfect hash family of size $O\left(\frac{d^2 \log n}{\log(q/e(d(d-1)/2+1))}\right)$ is achieved.
• When $q = \frac{d(d-1)}{2} + 1 + o(d^2)$, the construction yields a size of $O\left(\frac{d^6 \log n}{\log d}\right)$, which is within a factor of $d^4$ of the lower bound for $q = d + O(1)$.
• For $q \geq d^{2+\delta}$, the construction achieves size $O\left(\frac{d^2 \log n}{\log q}\right)$, matching the lower bound up to a factor of $d$.
• The paper provides a linear-time construction of $(w,r)$-cover-free families of size $O\left((ce)^w r^{w+1} \log n\right)$, which is nearly optimal under certain parameter regimes.
• For separating hash families, the construction achieves size $O\left(D_2 \log n\right)$ when $q \geq (D_2+1)^c$, matching the lower bound up to logarithmic factors.
• The constructions are globally explicit and run in linear time in the size of the output, improving over prior polynomial-time or non-explicit constructions.