[Paper Review] Asymptotic probabilities of extension properties and random $l$-colourable structures

Vera Koponen|arXiv (Cornell University)|Apr 11, 2012

Advanced Topology and Set Theory4 citations

TL;DR

This paper establishes conditions under which extension axioms almost surely hold in random finite $l$-colourable structures, proving a zero-one law for such structures under both uniform and dimension-conditional probability measures. It identifies a dichotomy for asymptotic probabilities in classes with forbidden substructures and extends these results to $l$-colourable structures using combinatorial and probabilistic techniques.

ABSTRACT

We consider a set $\\mbK = \\bigcup_{n \\in \\mbbN}\\mbK_n$ of {\\em finite} structures such that all members of $\\mbK_n$ have the same universe, the cardinality of which approaches $\\infty$ as $n\ o\\infty$. Each structure in $\\mbK$ may have a nontrivial underlying pregeometry and on each $\\mbK_n$ we consider a probability measure, either the uniform measure, or what we call the {\\em dimension conditional measure}. The main questions are: What conditions imply that for every extension axiom $\\varphi$, compatible with the defining properties of $\\mbK$, the probability that $\\varphi$ is true in a member of $\\mbK_n$ approaches 1 as $n \ o \\infty$? And what conditions imply that this is not the case, possibly in the strong sense that the mentioned probability approaches 0 for some $\\varphi$? If each $\\mbK_n$ is the set of structures with universe ${1, ..., n}$, in a fixed relational language, in which certain "forbidden" structures cannot be weakly embedded and $\\mbK$ has the disjoint amalgamation property, then there is a condition (concerning the set of forbidden structures) which, if we consider the uniform measure, gives a dichotomy; i.e. the condition holds if and only if the answer to the first question is `yes'. In general, we do not obtain a dichotomy, but we do obtain a condition guaranteeing that the answer is `yes' for the first question, as well as a condition guaranteeing that the answer is `no'; and we give examples showing that in the gap between these conditions the answer may be either `yes' or `no'. This analysis is made for both the uniform measure and for the dimension conditional measure. The later measure has closer relation to random generation of structures and is more "generous" with respect to satisfiability of extension axioms.

Motivation & Objective

• To determine conditions under which extension axioms almost surely hold in finite $l$-colourable structures as size grows.
• To identify dividing lines—conditions that distinguish whether extension axioms hold with probability approaching 1 or 0.
• To extend zero-one laws from general finite structures to the specific class of $l$-colourable structures.
• To compare the behavior of two probability measures: uniform and dimension-conditional, in relation to extension axiom satisfaction.
• To analyze the role of pregeometries and closure operators in shaping asymptotic probabilities of logical properties.

Proposed method

• Uses a framework of classes $\mathbf{K} = \bigcup_{n\in\mathbb{N}} \mathbf{K}_n$ of finite $L$-structures with increasing universe size.
• Applies two probability measures: uniform measure $\mu_n(\mathcal{M}) = 1/|\mathbf{K}_n|$ and dimension-conditional measure based on closure operators.
• Introduces $k$-extension axioms as logical statements expressing embeddability of substructures under closure constraints.
• Employs combinatorial estimates on multichromatic tuples: $\overline{\mathrm{smult}}(n,\gamma,m)$ counts $m$-tuples with distinct colours under colouring $\gamma$.
• Uses inequalities to compare $\overline{\mathrm{smult}}(n,\sigma_n,m)$ for balanced colourings $\sigma_n$ versus unbalanced $\gamma_n$, proving a lower bound difference of order $\Omega(n^m)$.
• Applies the probabilistic method to show that the fraction of $l$-colourable structures failing to satisfy certain extension axioms vanishes as $n \to \infty$.

Experimental results

Research questions

• RQ1 Under what conditions does the probability that a given extension axiom holds in a random structure from $\mathbf{K}_n$ approach 1 as $n \to \infty$?
• RQ2 When does the probability of an extension axiom approach 0, especially in classes with forbidden substructures?
• RQ3 Can a zero-one law be established for random $l$-colourable structures under the uniform and dimension-conditional measures?
• RQ4 How do pregeometries and closure operators affect the asymptotic behavior of extension axioms?
• RQ5 What is the role of colouring richness (e.g., $\frac{n}{a}$-rich) in determining the likelihood of realizing extension properties?

Key findings

• For classes with the disjoint amalgamation property and forbidden substructures, a dichotomy exists: extension axioms hold almost surely if and only if a specific condition on forbidden substructures is satisfied.
• The dimension-conditional measure is more permissive in realizing extension axioms than the uniform measure, due to its bias toward structures with higher dimension.
• In random $l$-colourable structures, a restricted set of extension axioms holds almost surely, enabling a zero-one law under both the dimension-conditional and uniform measures.
• The difference $\overline{\mathrm{smult}}(n,\sigma_n,m) - \overline{\mathrm{smult}}(n,\gamma_n,m)$ is bounded below by $\lambda_m n^m$ for some $\lambda_m > 0$, ensuring that balanced colourings dominate in multichromatic tuple counts.
• The probability that a random $l$-colourable structure fails to be $\frac{n}{a}$-rich for large $a$ tends to zero, which is key to proving the zero-one law.
• For $2 \leq m \leq l$, the inequality $\frac{1}{l^m}\binom{l}{m} > \frac{1}{(l-1)^m}\binom{l-1}{m}$ holds, ensuring a positive lower bound in the asymptotic difference of multichromatic tuple counts.