ABSTRACT. Building upon the theory of graph limits and the Aldous-Hoover representation and inspired by Panchenko's work on asymptotic Gibbs measures [Annals of Probability 2013], we construct continuous embeddings of discrete probability distributions. We show that the theory of graph limits induces a meaningful notion of convergence and derive a corresponding version of the Szemerédi regularity lemma. Moreover, complementing recent work , we apply these results to Gibbs measures induced by sparse random factor graphs and verify the "replica symmetric solution" predicted in the physics literature under the assumption of non-reconstruction.

No abstract

We establish a multivariate local limit theorem for the order and size as well as several other parameters of the k-core of the Erdős-Rényi random graph. The proof is based on a novel approach to the k-core problem that replaces the meticulous analysis of the 'peeling process' by a generative model of graphs with a core of a given order and size. The generative model, which is inspired by the Warning Propagation message passing algorithm, facilitates the direct study of properties of the core and its connections with the mantle and should therefore be of interest in its own right.The formula (1.5) determines the asymptotic probability that the order and size X , Y of the k-core attain specific values within O( n) of their expectations. Hence, Theorem 1.1 provides a bivariate local limit theorem for the order and size of the k-core. This result is significantly stronger than a mere central limit theorem stating that X , Y converge jointly to a bivariate Gaussian because (1.5) actually yields the asymptotic point probabilities. Still it is worthwhile pointing out that Theorem 1.1 immediately implies a central limit theorem. Corollary 1.2. Suppose that k ≥ 3 and d > d k , let Q be the matrix from (1.4) and let X , Y be the order and size of the k-core of G. Then n −1/2 ((X − np(1− q)), 2(Y − mp 2 )/d) converges in distribution to a bivariate Gaussian with mean 0 and covariance matrix Q −1 .A statement similar to Corollary 1.2 was previously established by Janson and Luczak [17] via a careful analysis of the peeling process. However, they did not obtain an explicit formula for the covariance matrix. Indeed, although the formula for Q is a bit on the lengthy side, the only non-algebraic quantity is p = p(d, k), the solution to the fixed point equation. By contrast, the formula of Janson and Luczak implicitly characterises the covariance matrix in terms of another stochastic process, and they do not provide a local limit theorem.The number d k from (1.2) does, of course, coincide with the k-core threshold first derived in [26]. The formula given in that paper looks a bit different but we pointed out the equivalence in [4]. In fact, it is very easy to show 2

ABSTRACT. The k-core, defined as the largest subgraph of minimum degree k, of the random graph G(n, p) has been studied extensively. In a landmark paper Pittel, Wormald and Spencer [Journal of Combinatorial Theory, Series B 67 (1996) 111-151] determined the threshold d k for the appearance of an extensive k-core. Here we derive a multi-type branching process that describes precisely how the k-core is "embedded" into the random graph for any k ≥ 3 and any fixed average degree d = np > d k . This generalises prior results on, e.g., the internal structure of the k-core.

No abstract