m‐ary Search trees when m ≥ 27: A strong asymptotics for the space requirements

It is known that the joint distribution of the number of nodes of each type of an m‐ary search tree is asymptotically multivariate normal when m ≤ 26. When m ≥ 27, we show the following strong asymptotics of the random vector X_n = ^t(X, … , X), where X denotes the number of nodes containing i ? 1 keys after having introduced n ? 1 keys in the tree: There exist (nonrandom) vectors X, C, and S and random variables ρ and φ such that (X_n ? nX)/n ? ρ(C cos(τ₂log n + φ) + S sin(τ₂log n + φ)) →_n→∞ 0 almost surely and in L²; σ₂ and τ₂ denote the real and imaginary parts of one of the eigenvalues of the transition matrix, having the second greatest real part. © 2004 Wiley Periodicals, Inc. Random Struct. Alg., 2004     

The distribution of nodes of given degree in random trees

Let 𝒯_n denote the set of unrooted unlabeled trees of size n and let k ≥ 1 be given. By assuming that every tree of 𝒯_n is equally likely, it is shown that the limiting distribution of the number of nodes of degree k is normal with mean value ∼ μ_kn and variance ∼ σn with positive constants μ_k and σ_k. Besides, the asymptotic behavior of μ_k and σ_k for k → ∞ as well as the corresponding multivariate distributions are derived. Furthermore, similar results can be proved for plane trees, for labeled trees, and for forests. © 1999 John Wiley & Sons, Inc. J Graph Theory 31: 227–253, 1999     

The distribution of the size of the ancestor‐tree and of the induced spanning subtree for random trees

We consider two tree statistics that extend in a natural way the parameters depth of a node resp. distance between two nodes. The ancestor‐tree of p given nodes in a rooted tree T is the subtree of T, spanned by the root and these p nodes and generalizes the depth (ancestor‐tree of a single node), whereas the spanning subtree induced by p given nodes in a tree T generalizes the distance (induced spanning subtree of two nodes). We study the random variables size of the ancestor‐tree resp. spanning subtree size for two tree families, the simply generated trees and the recursive trees. We will assume here the random tree model and also that all () possibilities of selecting p nodes in a tree of size n are equally likely. For random simply generated trees we can then characterize for a fixed number p of chosen nodes the limiting distribution of both parameters as generalized Gamma distributions, where we prove the convergence of the moments too. For some specific simply generated tree families we can give exact formulæ for the first moments. In the instance of random recursive trees, we will show that the considered parameters are asymptotically normally distributed, where we can give also exact formulæ for the expectation and the variance. © 2004 Wiley Periodicals, Inc. Random Struct. Alg., 2004     

The ζ(2) limit in the random assignment problem

The random assignment (or bipartite matching) problem asks about A_n=min_π ∑c(i, π(i)), where (c(i, j)) is a n×n matrix with i.i.d. entries, say with exponential(1) distribution, and the minimum is over permutations π. Mézard and Parisi (1987) used the replica method from statistical physics to argue nonrigorously that EA_n→ζ(2)=π²/6. Aldous (1992) identified the limit in terms of a matching problem on a limit infinite tree. Here we construct the optimal matching on the infinite tree. This yields a rigorous proof of the ζ(2) limit and of the conjectured limit distribution of edge‐costs and their rank‐orders in the optimal matching. It also yields the asymptotic essential uniqueness property: every almost‐optimal matching coincides with the optimal matching except on a small proportion of edges. ©2001 John Wiley & Sons, Inc. Random Struct. Alg., 18: 381–418, 2001     

Distances in random graphs with finite variance degrees

In this paper we study a random graph with N nodes, where node j has degree D_j and {D_j} are i.i.d. with ?(D_j ≤ x) = F(x). We assume that 1 ? F(x) ≤ cx^?τ+1 for some τ > 3 and some constant c > 0. This graph model is a variant of the so‐called configuration model, and includes heavy tail degrees with finite variance. The minimal number of edges between two arbitrary connected nodes, also known as the graph distance or the hopcount, is investigated when N → ∞. We prove that the graph distance grows like log_ν N, when the base of the logarithm equals ν = ??[D_j(D_j ? 1)]/??[D_j] > 1. This confirms the heuristic argument of Newman, Strogatz, and Watts [Phys Rev E 64 (2002), 026118, 1–17]. In addition, the random fluctuations around this asymptotic mean log_ν N are characterized and shown to be uniformly bounded. In particular, we show convergence in distribution of the centered graph distance along exponentially growing subsequences. © 2005 Wiley Periodicals, Inc. Random Struct. Alg., 2005     

Level of nodes in increasing trees revisited

Simply generated families of trees are described by the equation T(z) = ϕ(T(z)) for their generating function. If a tree has n nodes, we say that it is increasing if each node has a label ∈ { 1,…,n}, no label occurs twice, and whenever we proceed from the root to a leaf, the labels are increasing. This leads to the concept of simple families of increasing trees. Three such families are especially important: recursive trees, heap ordered trees, and binary increasing trees. They belong to the subclass of very simple families of increasing trees, which can be characterized in 3 different ways. This paper contains results about these families as well as about polynomial families (the function ϕ(u) is just a polynomial). The random variable of interest is the level of the node (labelled) j, in random trees of size n ≥ j. For very simple families, this is independent of n, and the limiting distribution is Gaussian. For polynomial families, we can prove this as well for j,n → ∞ such that n − j is fixed. Additional results are also given. These results follow from the study of certain trivariate generating functions and Hwang's quasi power theorem. They unify and extend earlier results by Devroye, Mahmoud, and others. © 2007 Wiley Periodicals, Inc. Random Struct. Alg., 2007     

Phase transition for Parking blocks,Brownian excursion and coalescence

In this paper, we consider hashing with linear probing for a hashing table with m places, n items (n < m), and ? = m ? n empty places. For a noncomputer science‐minded reader, we shall use the metaphore of n cars parking on m places: each car c_i chooses a place p_i at random, and if p_i is occupied, c_i tries successively p_i + 1, p_i + 2, until it finds an empty place. Pittel [42] proves that when ?/m goes to some positive limit β < 1, the size B of the largest block of consecutive cars satisfies 2(β ? 1 ? log β)B = 2 log m ? 3 log log m + Ξ_m, where Ξ_m converges weakly to an extreme‐value distribution. In this paper we examine at which level for n a phase transition occurs between B = o(m) and m ? B = o(m). The intermediate case reveals an interesting behavior of sizes of blocks, related to the standard additive coalescent in the same way as the sizes of connected components of the random graph are related to the multiplicative coalescent. © 2002 Wiley Periodicals, Inc. Random Struct. Alg., 21: 76–119, 2002     

Bounded branching process and and/or tree evaluation

We study the tail distribution of supercritical branching processes for which the number of offspring of an element is bounded. Given a supercritical branching process {Z_n} with a bounded offspring distribution, we derive a tight bound, decaying super-exponentially fast as c increases, on the probability Pr[Z_n > cE(Z_n)], and a similar bound on the probability Pr[Z_n ≤ E(Z_n)/c] under the assumption that each element generates at least two offspring. As an application, we observe that the execution of a canonical algorithm for evaluating uniform AND/OR trees in certain probabilistic models can be viewed as a two-type supercritical branching process with bounded offspring, and show that the execution time of this algorithm is likely to concentrate around its expectation, with a standard deviation of the same order as the expectation.     

Subgraphs smaller than the girth

For graphs A, B, let () denote the number of subsets of nodes of A for which the induced subgraph is B. If G and H both have girth > k, and if () = () for every k-node tree T, then for every k-node forest F, () = (). Say the spread of a tree is the number of nodes in a longest path. If G is regular of degree d, on n nodes, with girth > k, and if F is a forest of total spread ≤k, then the value of () depends only on n and d.     

The determining number of a Cartesian product

A set S of vertices is a determining set for a graph G if every automorphism of G is uniquely determined by its action on S. The determining number of G, denoted Det(G), is the size of a smallest determining set. This paper begins by proving that if G=G□?□G is the prime factor decomposition of a connected graph then Det(G)=max{Det(G)}. It then provides upper and lower bounds for the determining number of a Cartesian power of a prime connected graph. Further, this paper shows that Det(Q_n)=?log₂n?+1 which matches the lower bound, and that Det(K)=?log₃(2n+1)?+1 which for all n is within one of the upper bound. The paper concludes by proving that if H is prime and connected, Det(Hⁿ)=Θ(logn). © 2009 Wiley Periodicals, Inc. J Graph Theory     

On invariant measures of extensions of Markov transition functions

By using the LITTLEWOOD matrices A_2n we generalize CLARKSON' S inequalities, or equivalently, we determine the norms ‖A_2n: l(L_P) → l(L_P)‖ completely. The result is compared with the norms ‖A_2n: l → l‖, which are calculated implicitly in PIETSCH [6].     

Word maps and spectra of random graph lifts

We study here the spectra of random lifts of graphs. Let G be a finite connected graph, and let the infinite tree T be its universal cover space. If λ₁ and ρ are the spectral radii of G and T respectively, then, as shown by Friedman (Graphs Duke Math J 118 (2003), 19–35), in almost every n‐lift H of G, all “new” eigenvalues of H are ≤ O(λ ρ^1/2). Here we improve this bound to O(λ ρ^2/3). It is conjectured in (Friedman, Graphs Duke Math J 118 (2003) 19–35) that the statement holds with the bound ρ + o(1) which, if true, is tight by (Greenberg, PhD thesis, 1995). For G a bouquet with d/2 loops, our arguments yield a simple proof that almost every d‐regular graph has second eigenvalue O(d^2/3). For the bouquet, Friedman (2008). has famously proved the (nearly?) optimal bound of . Central to our work is a new analysis of formal words. Let w be a formal word in letters g,…,g. The word map associated with w maps the permutations σ₁,…,σ_k ∈ S_n to the permutation obtained by replacing for each i, every occurrence of g_i in w by σ_i. We investigate the random variable X that counts the fixed points in this permutation when the σ_i are selected uniformly at random. The analysis of the expectation ??(X) suggests a categorization of formal words which considerably extends the dichotomy of primitive vs. imprimitive words. A major ingredient of a our work is a second categorization of formal words with the same property. We establish some results and make a few conjectures about the relation between the two categorizations. These conjectures suggest a possible approach to (a slightly weaker version of) Friedman's conjecture. As an aside, we obtain a new conceptual and relatively simple proof of a theorem of A. Nica (Nica, Random Struct Algorithms 5 (1994), 703–730), which determines, for every fixed w, the limit distribution (as n →∞) of X. A surprising aspect of this theorem is that the answer depends only on the largest integer d so that w = u^d for some word u. © 2010 Wiley Periodicals, Inc. Random Struct. Alg., 2010     

An artificial boundary condition for an advection–diffusion equation

Consider the advection–diffusion equation: u₁ + au_x1 ? vδu = 0 in ?ⁿ × ?⁺ with initial data u⁰; the Support of u⁰ is contained in ?(x₁ < 0) and a: ?ⁿ → ? is positive. In order to approximate the full space solution by the solution of a problem in ? × ?⁺, we propose the artificial boundary condition: u₁ + au_x1 = 0 on ∑. We study this by means of a transmission problem: the error is an O(v²) for small values of the viscosity v.     

Markov chains on hypercubes: Spectral representations and several majorization relations

Various Markov chains on hypercubes ??are considered and their spectral representations are presentend in terms of Kronecker products. Special attention is given to random walks on the graphs ??(l = 1,n ? 2), where the vertex set is ?? and two vertices are connected if and only if their Hamming distance is at most l. It is shown that λ(??₁)>λ(??₁)>λ(??_n?1)>λ(??_n),l=2,…,n?2, where λ (??_I) is the specturum of the random walk on ??I, and > denotes the majorization ordering. A similar majorization relation is established for graphs V₁ where two veritces are connected if and only if their Hamming distance is exactly l. Some applications to mean times of these random walks are given. © 1993 John Wiley & Sons, Inc.     

Monotone numerical schemes for a Dirichlet problem for elliptic operators in divergence form

We consider a second‐order differential operator A( x )=?∑?_ia_ij( x )?_j+ ∑?_j(b_j( x )·)+c( x ) on ?^d, on a bounded domain D with Dirichlet boundary conditions on ?D, under mild assumptions on the coefficients of the diffusion tensor a_ij. The object is to construct monotone numerical schemes to approximate the solution of the problem A( x )u( x )=µ( x ), x ∈D, where µ is a positive Radon measure. We start by briefly mentioning questions of existence and uniqueness introducing function spaces needed to prove convergence results. Then, we define non‐standard stencils on grid‐knots that lead to extended discretization schemes by matrices possessing compartmental structure. We proceed to discretization of elliptic operators, starting with constant diffusion tensor and ending with operators in divergence form. Finally, we discuss W‐convergence in detail, and mention convergence in C and L₁ spaces. We conclude by a numerical example illustrating the schemes and convergence results. Copyright © 2008 John Wiley & Sons, Ltd.     

On restricted min‐wise independence of permutations

A family of permutations ℱ ⊆ S_n with a probability distribution on it is called k-restricted min-wise independent if we have Pr[min π(X) = π(x)] = 1/|X| for every subset X ⊆ [n] with |X| ≤ k, every x ∈ X, and π ∈ ℱ chosen at random. We present a simple proof of a result of Norin: every such family has size at least Some features of our method might be of independent interest. The best available upper bound for the size of such family is 1 + ∑ (j − 1)(). We show that this bound is tight if the goal is to imitate not the uniform distribution on S_n, but a distribution given by assigning suitable priorities to the elements of [n] (the stationary distribution of the Tsetlin library, or self-organizing lists). This is analogous to a result of Karloff and Mansour for k-wise independent random variables. We also investigate the cases where the min-wise independence condition is required only for sets X of size exactly k (where we have only an Ω(log log n + k) lower bound), or for sets of size k and k − 1 (where we already obtain a lower bound of n − k + 2). © 2003 Wiley Periodicals, Inc. Random Struct. Alg., 2003     

On the number of cycles of given length of a free word in several random permutations

Let w ≠ 1 be a free word in the symbols g₁,…, g_k and their inverses (i.e., an element of the free group F_k). For any s₁,…, s_k, in the group s_n of all permutation of n objects, we denote by w(s₁,…,s_k) ? S_n the permutation obtained by replacing g₁,…, g_k with s₁,…, s_k in the expression of w. Let X (s₁,…, s_k) denote the number of cycles of length L of w(s₁,…, s_k). For fixed w and L, we show that X, viewed as a random variable on S_n^k, has (for n →∞) a Poisson-type limit distribution, which can be computed precisely. © 1994 John Wiley & Sons, Inc.     

Iteratively reweighted least squares minimization for sparse recovery

Under certain conditions (known as the restricted isometry property, or RIP) on the m × N matrix Φ (where m < N), vectors x ∈ ?^N that are sparse (i.e., have most of their entries equal to 0) can be recovered exactly from y := Φx even though Φ^?1(y) is typically an (N ? m)—dimensional hyperplane; in addition, x is then equal to the element in Φ^?1(y) of minimal ??₁‐norm. This minimal element can be identified via linear programming algorithms. We study an alternative method of determining x, as the limit of an iteratively reweighted least squares (IRLS) algorithm. The main step of this IRLS finds, for a given weight vector w, the element in Φ^?1(y) with smallest ??₂(w)‐norm. If x⁽ⁿ⁾ is the solution at iteration step n, then the new weight w⁽ⁿ⁾ is defined by w := [|x|² + ε]^?1/2, i = 1, …, N, for a decreasing sequence of adaptively defined ε_n; this updated weight is then used to obtain x^{(n + 1)} and the process is repeated. We prove that when Φ satisfies the RIP conditions, the sequence x⁽ⁿ⁾ converges for all y, regardless of whether Φ^?1(y) contains a sparse vector. If there is a sparse vector in Φ^?1(y), then the limit is this sparse vector, and when x⁽ⁿ⁾ is sufficiently close to the limit, the remaining steps of the algorithm converge exponentially fast (linear convergence in the terminology of numerical optimization). The same algorithm with the “heavier” weight w = [|x|² + ε]^?1+τ/2, i = 1, …, N, where 0 < τ < 1, can recover sparse solutions as well; more importantly, we show its local convergence is superlinear and approaches a quadratic rate for τ approaching 0. © 2009 Wiley Periodicals, Inc.     

Coin flipping from a cosmic source: On error correction of truly random bits

We study a problem related to coin flipping, coding theory, and noise sensitivity. Consider a source of truly random bits x ∈ {0, 1}ⁿ, and k parties, who have noisy version of the source bits yⁱ ∈ {0, 1}ⁿ, when for all i and j, it holds that P [y = x_j] = 1 ? ?, independently for all i and j. That is, each party sees each bit correctly with probability 1 ? ?, and incorrectly (flipped) with probability ?, independently for all bits and all parties. The parties, who cannot communicate, wish to agree beforehand on balanced functions f_i: {0, 1}ⁿ → {0, 1} such that P [f₁(y¹) = … = f_k(y^k)] is maximized. In other words, each party wants to toss a fair coin so that the probability that all parties have the same coin is maximized. The function f_i may be thought of as an error correcting procedure for the source x. When k = 2,3, no error correction is possible, as the optimal protocol is given by f_i(yⁱ) = y. On the other hand, for large values of k, better protocols exist. We study general properties of the optimal protocols and the asymptotic behavior of the problem with respect to k, n, and ?. Our analysis uses tools from probability, discrete Fourier analysis, convexity, and discrete symmetrization. © 2005 Wiley Periodicals, Inc. Random Struct. Alg., 2005     

Limit laws for local counters in random binary search trees

Limit laws for several quantities in random binary search trees that are related to the local shape of a tree around each node can be obtained very simply by applying central limit theorems for w-dependent random variables. Examples include: the number of leaves (L_n), the number of nodes with k descendants (k fixed), the number of nodes with no left child, the number of nodes with k left descendants. Some of these results can also be obtained via the theory of urn models, but the present method seems easier to apply.