arXiv++ Combinatorics

Relation between the real eigenvalues of the non-backtracking matrix and those of the non-backtracking Laplacian is considered with respect to node clustering. For this purpose we use the real eigenvalues of the transition probability matrix (when the random walk goes through the oriented edges with the rule of ``not going back in the next step'') which have a linear relation to those of the non-backtracking Laplacian of Jost and Mulas. ``Inflation--deflation'' techniques are also developed for clustering the nodes of the non-backtracking graph. With further processing, it leads to the clustering of the nodes of the original graph, which usually comes from a sparse stochastic block model of Bordenave and Decelle.

We solve two long standing open problems, one from probability theory formulated by Malyshev in 1970 and another one from a crossroad of geometry and dynamics, of Darboux from 1879. The Malyshev problem is of finding effective, explicit necessary and sufficient conditions in the closed form to characterize all random walks in the quarter plane with the finite group of random walk of order $2n$, for all $n\ge 2$, where the underlining biquadratic is an elliptic curve. Until now, the results were known only for $n=2, 3, 4$, obtained using ad-hoc methods developed separately for each of the three cases. We provide a method that solves the problem for all $n$ and in a unified way. Explicit examples of random walks with the groups of orders higher than 10 are presented here for the first time, including orders 12, 14, 16. The same method applies to any higher order. We consider cases with singular biquadratics in a systematic manner. We establish a new two-way relationship between diagonal random walks and $4$-bar links. We describe all $n$-periodic Darboux transformations for $4$-bar links for all $n\ge 2$, thus completely solving the Darboux problem: after $n$ iterations, a polygonal configuration maps to a congruent one of the same orientation, that he solved for $n=2$, which was recently extended to $n=3$. We also study $k$-semi-periodicity as a natural type of periodicity of the Darboux transformations, where after $k$ iterations of the Darboux transformation, a polygonal configuration maps to a congruent one, but of opposite orientation. By introducing a new object, the secondary $(2,2)$ correspondence, and the related secondary cubic of the centrally-symmetric biquadratics, we provide necessary and sufficient conditions for $k$-semi-periodicity for $4$-bar links for all $k\ge 2$ in an explicit closed form, while the case $k=2$ was solved recently.

In this paper we study the number of returns to the coordinate hyperplanes for multidimensional nearest-neighbour random walks. While one-dimensional results on returns are classical, much less is known in higher dimensions. We analyse the asymptotic behaviour of returns under several natural conditionings: the unconditioned walk, bridges, meanders, and non-negative bridges (or excursions). Our main results characterize the limiting distributions under appropriate rescaling. The resulting one-dimensional marginals may be half-normal, Rayleigh, geometric, negative binomial, or certain mixtures thereof. In most situations, the coordinates are asymptotically independent; however, there are notable exceptions for the meander case, depending on the drift. The proofs rely on conditioning on the numbers of horizontal and vertical steps, which restores a form of independence and reduces the problem to one-dimensional estimates via binomial convolution and Bernstein-type approximations.

We show that random walks on semisimple algebraic groups do not concentrate on proper algebraic subvarieties with uniform exponential rate of anti-concentration. This is achieved by proving a uniform spectral gap for quasi-regular representations of countable linear groups. The method makes key use of Diophantine heights and the Height Gap theorem. We also deduce a non-abelian version of the Littlewood--Offord inequalities and prove logarithmic bounds for escape from subvarieties. In a sequel to this paper, we will show how to transform this uniform gap into uniform expansion for Cayley graphs of finite simple groups of bounded rank $G(p)$ over almost all primes $p$.

We use representation theory of $S_n$ to analyze the mixing of permutation cycle type statistics $a_j(σ) = ${# of $j$-cycles of $σ$} for any fixed $j$ and $σ$ resulting from a random $i$-cycle walk on $S_n$. We also derive analogous results for the random star transposition walk. Our approach uses the method of moments; a key ingredient is a new formula for the coefficients in the irreducible character decomposition of the $S_n$-class function $(a_j)^r(σ)=\{(\text{# of $j$-cycles of $σ$})^r\}$ for any positive integers $r,j$ when $n\geq 2rj$.

We define the Uniform Random Walk (URW) on a connected, locally finite graph as the weak limit of the uniform walk of length $n$ starting at a fixed vertex. When the limit exists, it is necessarily Markovian and is independent of the starting point. For a finite graph, URW equals the Maximal Entropy Random Walk (MERW). We investigate the existence and phase transitions of URW for loop perturbed regular graphs and their limits. It turns out that for a sequence of finite graphs, it is the global spectral theory of the limiting graphing that governs the behavior of the finite MERWs. In the delocalized phase, we use a "membrane argument", showing that the principal eigenfunction of an expander graphing is stable under a small diagonal perturbation. This gives us: 1) The existence of URW on leaves; 2) The URW is a unique entropy maximizer; 3) The MERW of a finite graph sequence Benjamini-Schramm converges to the URW of the limiting graphing. In the localized phase, the environment seen by the particle takes the role of a finite stationary measure. We show that for canopy trees, the URW exists, is transient and maximizes entropy. We also show that for large finite graphs where most vertices have a fixed degree, localization of MERW is governed by the adjacency norm.

We study the geometric properties of graphs with non-negative Ollivier-Ricci curvature, a discrete analogue of non-negative Ricci curvature in Riemannian geometry. We prove that for each $d<\infty$ there exists a constant $C_d$ such that if $G=(V,E)$ is a finite graph with non-negative Ollivier-Ricci curvature and with degrees bounded by $d$ then the average log-volume growth and random walk displacement satisfy \[ \frac{1}{|V|} \sum_{x\in V} \log \#B(x,r) \leq \exp\left[C_d \sqrt{\log r}\right] = r^{o(1)} \] and \[ \frac{1}{|V|} \sum_{x\in V} \mathbf{E}_x [d(X_0,X_n)^2] \leq n \exp\left[C_d \sqrt{\log n}\right] = n^{1+o(1)} \] for every $n,r\geq 2$. This significantly strengthens a result of Salez (GAFA 2022), who proved that the average displacement of the random walk is $o(n)$ and deduced that non-negatively curved graphs of bounded degree cannot be expanders. Our results also apply to infinite transitive graphs and, more generally, to bounded-degree unimodular random rooted graphs of non-negative Ollivier-Ricci curvature.

We study the asymptotic size of decompositions of tensor powers of tilting modules for quantum groups (mostly at a complex root of unity). In type A1 we obtain a sharp result for the number of indecomposable summands, explained by a one dimensional half-line random walk with a periodic congruence constraint. In general type we prove a universal law: the dominant part is governed only by the dimension of the module, while the correction depends only on the root system, so the asymptotic size is largely independent of the specific tilting module.

The spectral gap theorem of Caputo, Liggett, and Richthammer states that on any connected graph equipped with edge weights, the 2nd eigenvalue of the interchange process equals the 2nd eigenvalue of the random walk process. In this work we characterize the 2nd eigenspace of the interchange process. We prove that this eigenspace is uniquely determined by the 2nd eigenvectors of the random walk process on every connected weighted graph except the $4$-cycle with uniform edge weights. The key to our proof is an induction scheme on the number of vertices, and involves the octopus (in)equality, representation theoretic computations, and graph Laplacian computations.

We consider random walks on a tree $G=(V,E)$ with stationary distribution $π_v = \mathrm{deg}(v)/2|E|$ for $v \in V$. Let the hitting time $H(v,w)$ denote the expected number of steps required for the random walk started at vertex $v$ to reach vertex $w$. We characterize the extremal tree structures for the best meeting time $T_{\mathrm{bestmeet}}(G) = \min_{w \in V} \sum_{v \in V} π_v H(v,w)$ for trees of order $n$ with diameter $d$. The best meeting time is maximized by the balanced double broom graph, and it is minimized by the balanced lever graph.

Banach's matchbox problem considers the setting of two matchboxes that each initially contain the same number of matches. Boxes are chosen with equal probability and a match removed each time. The problem concerns the law of the number of matches remaining in one box once the other box empties. Knuth considered a generalization of this problem whereby `big-choosers' arrive with probability $p$ and remove a match from the box with the most number remaining, and `little-choosers' arrive with probability $1-p$ and remove a match from the box with the least number remaining. In this paper we consider Knuth's generalization for the case of $k$ matchboxes. We determine the generating function for the expected number of matches remaining in $k-1$ matchboxes once a box first empties, a quantity we refer to as the `residue'. Interestingly, this generating function is a quotient whose denominator contains a generating function for a special case of the Raney numbers. The form for this generating function allows us to give an expression for the expected residue in terms of a sum that involves diagonal state return probabilities, where a diagonal state is a configuration in which all matchboxes each contain the same number of matches. We use analytic techniques to determine the asymptotic behaviour of this expected value for all values of $p$, which involves the study of an asymmetric random walk. We also consider the expected value of the order of the first return to a diagonal state and determine its asymptotic behaviour. The coefficients of the diagonal state probability generating function are shown to be related to `manila folder configurations in a filing cabinet', and we make this connection precise. This allows us to use known results for the enumeration of such manila folder configurations to give a closed form expression for the diagonal state return probabilities.

In this paper we study the mixing time of the simple random walk on the giant component of supercritical $d$-dimensional random geometric graphs generated by the unit intensity Poisson Point Process in a $d$-dimensional cube of volume $n$. With $r_g$ denoting the threshold for having a giant component, we show that for every $ε> 0$ and any $r \ge (1+ε)r_g$, the mixing time of the giant component is with high probability $Θ(n^{2/d}/r^{2})$, thereby closing a gap in the literature. The main tool is an isoperimetric inequality which holds, w.h.p., for any large enough vertex set, a result which we believe is of independent interest. Our analysis also implies that the relaxation time is of the same order.

We present practical algorithms for generating universal cycles uniformly at random. In particular, we consider universal cycles for shorthand permutations, subsets and multiset permutations, weak orders, and orientable sequences. Additionally, we consider de Bruijn sequences, weight-range de Bruin sequences, and de Bruijn sequences, with forbidden $0^z$ substring. Each algorithm, seeded with a random element from the given set, applies a random walk of an underlying Eulerian de Bruijn graph to obtain a random arborescence (spanning in-tree). Given the random arborescence and the de Bruijn graph, a corresponding random universal cycle can be generated in constant time per symbol. We present experimental results on the average cover time needed to compute a random arborescence for each object using a Las Vegas algorithm.

We consider bond percolation on the $d$-dimensional binary hypercube with $p=c/d$ for fixed $c>1$. We prove that the typical diameter of the giant component $L_1$ is of order $Θ(d)$, and the typical mixing time of the lazy random walk on $L_1$ is of order $Θ(d^2)$. This resolves long-standing open problems of Bollobás, Kohayakawa and Łuczak from 1994, and of Benjamini and Mossel from 2003. A key component in our approach is a new tight large deviation estimate on the number of vertices in $L_1$ whose proof includes several novel ingredients: a structural description of the residue outside the giant component after sprinkling, a tight quantitative estimate on the spread of the giant in the hypercube, and a stability principle which rules out the disintegration of large connected sets under thinning. This toolkit further allows us to obtain optimal bounds on the expansion in $L_1$.

We propose two possible definitions for a version of Kemeny's constant of a graph based on non-backtracking random walks (in place of the usual simple random walk). We show that these two definitions coincide for edge-transitive graphs, and give a condition generalizing edge-transitive for which equality holds, and investigate by how much they can differ in general. We compute our non-backtracking Kemeny's constant for several families of graphs.

We investigate a randomly evolving process of subgraphs in an underlying host graph using the spectral theory of semigroups related to the Tsetlin library and hyperplane arrangements. Starting with some initial subgraph, at each iteration, we apply a randomly selected edit to the current subgraph. Such edits vary in nature from simple edits consisting of adding or deleting an edge, or compound edits which can affect several edges at once. This evolving process generates a random walk on the set of all possible subgraphs of the host graph. We show that the eigenvalues of this random walk can be naturally indexed by subsets of edges of the host graph. We also provide, in the case of simple edits, a closed-form formula for the eigenvectors of the transition probability matrix and a sharp bound for the rate of convergence of this random walk. We consider extensions to the case of compound edits; examples of this model include the previously studied Moran forest model and a dynamic random intersection graph model. Evolving graphs arise in a variety of fields ranging from deep learning and graph neural networks to epidemic modeling and social networks. Our random evolving process serves as a general stochastic model for sampling random subgraphs from a given graph.

Consider a random walk on a tree $G=(V,E)$. For $v,w \in V$, let the hitting time $H(v,w)$ denote the expected number of steps required for the random walk started at $v$ to reach $w$, and let $π_v = \mathrm{deg}(v)/2|E|$ denote the stationary distribution for the random walk. We characterize the extremal tree structures for the meeting time $T_{\mathrm{meet}}(G) = \max_{w \in V} \sum_{v \in V} π_v H(v,w)$. For fixed order $n$ and diameter $d$, the meeting time is maximized by the broom graph. The meeting time is minimized by the balanced double broom graph, or a slight variant, depending on the relative parities of $n$ and $d$.

We study the limiting behavior of a Bienayme-Galton-Watson tree conditioned to have a large number of vertices and either a fixed number of leaves or a fixed number of internal nodes. The first biconditioning gives a universal result with respect to the offspring distribution. In contrast, the second case leads to a variety of limiting behaviors, ranging from condensation phenomena to more elongated tree structures, depending on the properties of the offspring distribution. To prove these results, we use tools from conditioned random walk theory and from analytic combinatorics.

Let $G=(V,E)$ be a finite, connected graph. We investigate a notion of boundary $\partial G \subseteq V$ and argue that it is well behaved from the point of view of potential theory. This is done by proving a number of discrete analogous of classical results for compact domains $Ω\subset \mathbb{R}^d$. These include (1) an analogue of Pólya's result that a random walk in $Ω$ typically hits the boundary $\partial Ω$ within $\lesssim \mbox{diam}(Ω)^2$ units of time, (2) an analogue of the Faber-Krahn inequality, (3) an analogue of the Hardy inequality, (4) an analogue of the Alexandrov-Bakelman-Pucci estimate, (5) a stability estimate for hot spots and (6) a Theorem of Björck stating that probability measures $μ$ that maximize $\int_{Ω\times Ω} \|x-y\|^α dμ(x) dμ(y)$ are fully supported in the boundary.

A pair of random walks $(R,S)$ on the vertices of a graph $G$ is {\it successful} if two tokens can be scheduled (moving only one token at a time) to travel along $R$ and $S$ without colliding. We consider questions related to P. Winkler's {\it clairvoyant demon problem}, which asks whether for random walks $R$ and $S$ on $G$, $Pr[\ (R,S) \mbox{ is successful }] >0$. We introduce the notion of an {\it evasive} walk on $G$: a walk $S$ so that for a random walk $R$ on $G$, $Pr[\ (R,S) \mbox{ is successful }]>0$. We characterize graphs $G$ having evasive walks, giving explicit constructions on such $G$. On a cycle, we show that with high probability the tokens must collide quickly. Finally we consider two variants of the problem for which, under certain assumptions on the graph $G$, we provide algorithms that schedule $(R,S)$ successfully with positive probability.