Sidorenko-Type Inequalities for Pairs of Trees

📅 2023-05-25
🏛️ arXiv.org
📈 Citations: 1
Influential: 0
📄 PDF
🤖 AI Summary
This paper investigates Sidorenko-type inequalities $ H succcurlyeq T $ between tree graphs, requiring $ t(H,G)^{|E(T)|} ge t(T,G)^{|E(H)|} $ for all graphs $ G $, where $ t(cdot,cdot) $ denotes homomorphism density. Methodologically, the work integrates information-theoretic arguments, the Kopparty–Rossman linear programming framework, homomorphism density analysis, and structural induction on forests. The contributions are threefold: (i) it establishes the first necessary and sufficient condition for $ H succcurlyeq T $ when both $ H $ and $ T $ are trees; (ii) it fully classifies all such pairs among trees with at most eight vertices; and (iii) it resolves the Csikvári–Lin problem completely, including a precise characterization of the star $ S_k $ and the 4-vertex path $ P_4 $ within the Sidorenko preorder. The results unify and generalize classical theorems by Leontovich and Sidorenko, and provide a complete description of the Sidorenko order on small trees.
📝 Abstract
Given two non-empty graphs $H$ and $T$, write $Hsucccurlyeq T$ to mean that $t(H,G)^{|E(T)|}geq t(T,G)^{|E(H)|}$ for every graph $G$, where $t(cdot,cdot)$ is the homomorphism density function. We obtain various necessary and sufficient conditions for two trees $H$ and $T$ to satisfy $Hsucccurlyeq T$ and determine all such pairs on at most 8 vertices. This extends results of Leontovich and Sidorenko from the 1980s and 90s. Our approach applies an information-theoretic technique to reduce the problem of showing that $Hsucccurlyeq T$ for two forests $H$ and $T$ to solving a linear program of Kopparty and Rossman. We also characterize trees $H$ which satisfy $Hsucccurlyeq S_k$ or $Hsucccurlyeq P_4$, where $S_k$ is the $k$-vertex star and $P_4$ is the $4$-vertex path and resolve a problem of Csikv'ari and Lin.
Problem

Research questions and friction points this paper is trying to address.

Sidorenko's inequality
tree comparison
mathematical relationships
Innovation

Methods, ideas, or system contributions that make the work stand out.

Information Theory
Linear Programming Transformation
Extending Leontovich and Sidorenko's Research
🔎 Similar Papers
No similar papers found.
N
Natalie C. Behague
Mathematics Institute, University of Warwick, Coventry, UK
G
Gabriel Crudele
Department of Mathematics and Statistics, McGill University, Montréal, Canada
Jonathan A. Noel
Jonathan A. Noel
University of Victoria
Extremal CombinatoricsProbabilistic CombinatoricsGraph TheoryGraph ColouringGraph Limits
L
Lina M. Simbaqueba
Department of Mathematics and Statistics, University of Victoria, Victoria, Canada