Skip to main content
QUICK REVIEW

[Paper Review] Ahlswede-Khachatrian Theorems: Weighted, Infinite, and Hamming

Yuval Filmus|arXiv (Cornell University)|Oct 3, 2016
Limits and Structures in Graph Theory10 references3 citations
TL;DR

This paper extends the Ahlswede–Khachatrian theorem to weighted, infinite, and Hamming settings by reworking the original proofs in the $μ_p$-measure framework. It fully characterizes the maximum $μ_p$-measure of $t$-intersecting families for all $p \in (0,1)$, including $p > 1/2$, and identifies all optimal families, resolving gaps in prior work that failed to describe optimal families or only applied for $p < 1/2$. The results unify and generalize classical extremal combinatorics across multiple domains.

ABSTRACT

The seminal complete intersection theorem of Ahlswede and Khachatrian gives the maximum cardinality of a $k$-uniform $t$-intersecting family on $n$ points, and describes all optimal families. We extend this theorem to several other settings: the weighted case, the case of infinitely many points, and the Hamming scheme. The weighted Ahlswede-Khachatrian theorem gives the maximal $μ_p$ measure of a $t$-intersecting family on $n$ points, where $μ_p(A) = p^{|A|} (1-p)^{n-|A|}$. As has been observed by Ahlswede and Khachatrian and by Dinur and Safra, this theorem can be derived from the classical one by a simple reduction. However, this reduction fails to identify the optimal families, and only works for $p &lt; 1/2$. We translate the two original proofs of Ahlswede and Khachatrian to the weighted case, thus identifying the optimal families in all cases. We also extend the theorem to the case $p &gt; 1/2$, using a different technique of Ahlswede and Khachatrian (the case $p = 1/2$ is Katona's intersection theorem). We then extend the weighted Ahlswede-Khachatrian theorem to the case of infinitely many points. The Ahlswede-Khachatrian theorem on the Hamming scheme gives the maximum cardinality of a subset of $\mathbb{Z}_m^n$ in which any two elements $x,y$ have $t$ positions $i_1,\ldots,i_t$ such that $x_{i_j} - y_{i_j} \in \{-(s-1),\ldots,s-1\}$. We show that this case corresponds to $μ_p$ with $p = s/m$, extending work of Ahlswede and Khachatrian, who considered the case $s = 1$. We also determine the maximum cardinality families. We obtain similar results for subsets of $[0,1]^n$, though in this case we are not able to identify all maximum cardinality families.

Motivation & Objective

  • To complete the characterization of maximum $μ_p$-measure $t$-intersecting families on $n$ points, including for $p > 1/2$, where prior methods failed.
  • To identify all optimal families in the weighted setting, resolving the lack of structural description in earlier reductions.
  • To extend the Ahlswede–Khachatrian theorem to the case of infinitely many points, defining $w_{\infty}(p,t)$ as the supremum $μ_p$-measure of $t$-intersecting families in $\{0,1\}^{\aleph_0}$.
  • To generalize the theorem to the Hamming scheme $\mathbb{Z}_m^n$, showing it corresponds to $\u03bc_p$ with $p = s/m$, and determine maximum families.
  • To provide a simplified, unified proof of the Erdős–Ko–Rado theorem in the $μ_p$ setting using a generalized circle argument.

Proposed method

  • Rephrasing the two original proofs of Ahlswede and Khachatrian (1997) in the weighted $μ_p$-measure setting to preserve structural information about optimal families.
  • Using a lifting argument via random restrictions to transfer results from finite to infinite settings, showing that $w_{\infty}(p,t) = \sup_n w(n,t,p)$.
  • Applying Hoeffding’s inequality and measure concentration to bound the $μ_p$-measure of families under $p$-biased sampling, enabling comparison with optimal families.
  • Extending Katona’s circle method to the $μ_p$ setting to give a concise proof of the Erdős–Ko–Rado theorem in the weighted case.
  • Defining and analyzing the family $\mathcal{F}_{t,r} = \{S : |S \cap [t+2r]| \geq t+r\}$ in the $μ_p$-measure framework to determine optimal thresholds for $p$.
  • Using permutation invariance and minimality arguments to prove uniqueness of optimal families under $p$-optimality, especially when $p$ is not a critical fraction $r/(t+2r-1)$.

Experimental results

Research questions

  • RQ1What is the maximum $μ_p$-measure of a $t$-intersecting family on $n$ points for all $p \in (0,1)$, including $p > 1/2$?
  • RQ2Which families achieve this maximum measure, and how can they be fully characterized in the weighted setting?
  • RQ3What is the supremum $μ_p$-measure of a $t$-intersecting family on infinitely many points, and how does it relate to the finite case?
  • RQ4How does the Ahlswede–Khachatrian theorem extend to the Hamming scheme $\mathbb{Z}_m^n$, and what is the correspondence with $\u03bc_p$ for $p = s/m$?
  • RQ5Can Katona’s circle argument be generalized to the $μ_p$ setting to yield a simpler proof of the Erdős–Ko–Rado theorem?

Key findings

  • The paper fully characterizes $w(n,t,p)$, the maximum $μ_p$-measure of a $t$-intersecting family on $n$ points, for all $n$, $t$, and $p \in (0,1)$, resolving the incomplete picture from prior work.
  • For $p \in \left[\frac{r}{t+2r-1}, \frac{r+1}{t+2r+1}\right]$, the maximum measure is $\mu_p(\mathcal{F}_{t,r})$, and this family is optimal for all $n \geq t+2r$, with explicit identification of all optimal families.
  • The case $p > 1/2$ is resolved using a different technique of Ahlswede and Khachatrian, showing that the optimal families remain of the form $\mathcal{F}_{t,r}$, but with $r$ chosen based on $p$.
  • The infinite case yields $w_{\infty}(t,p) = \sup_n w(n,t,p)$, and the supremum is achieved in the limit, with optimal families lifted from finite constructions.
  • In the Hamming scheme $\mathbb{Z}_m^n$, the maximum $t$-intersecting families correspond to $\u03bc_p$ with $p = s/m$, and the optimal families are again $\mathcal{F}_{t,r}$ for appropriate $r$, with explicit bounds on $n$.
  • A generalized circle argument in the $μ_p$ setting provides a short, elegant proof of the Erdős–Ko–Rado theorem, simplifying prior proofs by Dinur–Friedgut and Friedgut.

Better researchstarts right now

From reading papers to final review, dramatically reduce your research time.

No credit card · Free plan available

This review was created by AI and reviewed by human editors.