[Paper Review] Local Improvement Gives Better Expanders

Michael Lampis|arXiv (Cornell University)|Nov 2, 2012

Advanced Graph Theory Research11 references3 citations

TL;DR

This paper improves the lower bounds on the expansion of random Δ-regular graphs by introducing a local improvement argument: instead of analyzing all small-expansion sets, it focuses only on those that are locally optimal (i.e., no vertex swap reduces expansion). This refinement leads to significantly smaller probabilities for such sets to exist, enabling tighter union bound estimates. The key result is a strictly improved asymptotic lower bound on expansion, showing the classical bound by Bollobás is not tight and proving the existence of a constant α < 2√ln 2 such that random Δ-regular graphs have expansion at least (1 − α/√Δ)(Δ/2) with high probability.

ABSTRACT

It has long been known that random regular graphs are with high probability good expanders. This was first established in the 1980s by Bollobás by directly calculating the probability that a set of vertices has small expansion and then applying the union bound. In this paper we improve on this analysis by relying on a simple high-level observation: if a graph contains a set of vertices with small expansion then it must also contain such a set of vertices that is locally optimal, that is, a set whose expansion cannot be made smaller by exchanging a vertex from the set with one from the set's complement. We show that the probability that a set of vertices satisfies this additional property is significantly smaller. Thus, after again applying the union bound, we obtain improved lower bounds on the expansion of random $Δ$-regular graphs for $Δ\ge 4$. In fact, the gains from this analysis increase as $Δ$ grows, a fact we explain by extending our technique to general $Δ$. Thus, in the end we obtain an improvement not only for some small special cases but on the general asymptotic bound on the expansion of $Δ$-regular graphs given by Bollobás.

Motivation & Objective

• To improve the known lower bounds on the expansion of random Δ-regular graphs for small Δ ≥ 4.
• To show that the classical asymptotic lower bound of Δ/2 − Θ(√Δ) by Bollobás is not tight.
• To develop a refined analysis technique that leverages local optimality of small-expansion sets to reduce the probability of their existence.
• To demonstrate that the improvement is not ad hoc but generalizable across all Δ, including the asymptotic regime.
• To address open problems in inapproximability of bounded-occurrence CSPs by improving known expansion thresholds.

Proposed method

• Introduce the concept of locally optimal sets: sets whose expansion cannot be reduced by swapping a vertex in the set with one in the complement.
• Show that if any small-expansion set exists, then a locally optimal one must also exist, so it suffices to bound the probability of such sets.
• Use the union bound over locally optimal sets, which have significantly lower probability of occurrence than arbitrary small-expansion sets.
• Apply probabilistic analysis using degree distribution approximations and normal approximations to bound the probability of existence of such sets.
• Derive a new asymptotic lower bound on expansion by optimizing over parameters like η and θ that control the deviation of binomial variables.
• Use logarithmic expansions and tail bounds to refine the estimate of the logarithmic probability of existence of locally optimal sets.

Experimental results

Research questions

• RQ1 Can the expansion lower bound for random Δ-regular graphs be improved beyond the classical result of Bollobás (1988)?
• RQ2 Is the asymptotic bound of Δ/2 − Θ(√Δ) tight, or can the coefficient of √Δ be improved?
• RQ3 Does the local optimality condition on small-expansion sets lead to a significant reduction in the probability of their existence, enabling tighter union bound estimates?
• RQ4 Can the improved analysis be generalized to all Δ, not just small values, to yield a better asymptotic bound?
• RQ5 Do locally optimal sets provide a general-purpose tool for refining probabilistic bounds in random regular graphs?

Key findings

• The paper establishes a new asymptotic lower bound on expansion: for any Δ ≥ 4, random Δ-regular graphs have expansion at least (1 − α/√Δ)(Δ/2) with high probability, where α < 2√ln 2 ≈ 1.66.
• The improvement is not limited to small Δ; the gains increase with Δ, showing the classical bound is not tight asymptotically.
• For small Δ, such as Δ = 6, the bound implies expansion at least 0.75Δ, improving on previous estimates.
• The probability that a locally optimal set of size d exists is significantly smaller than for arbitrary sets, which reduces the union bound contribution.
• The analysis shows that the expected number of locally optimal sets with expansion below a threshold is o(1), implying such sets exist with low probability.
• Numerical evidence and asymptotic analysis suggest that the optimal local improvement condition occurs when d = d′, though a general proof remains open.