On a question of Vera T. S\'os about size forcing of graphons

TL;DR

This paper investigates whether the distribution of the number of edges in random samples from a graphon uniquely determines the graphon, focusing on 2-step graphons and providing positive results for a specific family.

Contribution

It extends previous work by analyzing size forcing for 2-step graphons, proving uniqueness for a 3-dimensional family, and presenting related findings.

Findings

Positive answer for a 3-dimensional family of 2-step graphons

Extension of size forcing results beyond constant graphons

Insights into the uniqueness of graphons from edge distribution data

Abstract

The $k$-sample $\mathbb{G}(k,W)$ from a graphon $W:[0,1]^2\to [0,1]$ is the random graph on $\{1,\dots,k\}$, where we sample $x_1,\dots,x_k\in [0,1]$ uniformly at random and make each pair $\{i,j\}\subseteq \{1,\dots,k\}$ an edge with probability $W(x_i,x_j)$, with all these choices being mutually independent. Let the random variable $X_k(W)$ be the number of edges in $\mathbb{G}(k,W)$. Vera T. S\'os asked in 2012 whether two graphons $U,W$ are necessarily weakly isomorphic if the random variables $X_k(U)$ and $X_k(W)$ have the same distribution for every integer $k\ge 2$. This question when one of the graphons $W$ is a constant function was answered positively by Endre Cs\'oka and independently by Jacob Fox, Tomasz {\L}uczak and Vera T. S\'os. Here we investigate the question when $W$ is a 2-step graphon and prove that the answer is positive for a 3-dimensional family of such graphons. We also present some related results.