arXiv subjects
Publications and source records attributed to Tomas Juškevičius.
In this paper we give anticoncentration bounds for sums of independent random vectors in finite-dimensional vector spaces. In particular, we asymptotically establish a conjecture of Leader and Radcliffe (1994) and a question of Jones (1978). The highlight of this work is an application of the strong perfect graph theorem by Chudnovsky, Robertson, Seymour and Thomas (2003) in the context of anticoncentration.
Let $X$ be a random variable and define its concentration function by $$\mathcal{Q}_{h}(X)=\sup_{x\in \mathbb{R}}\mathbb{P}(X\in (x,x+h]).$$ For a sum $S_n=X_1+\cdots+X_n$ of independent real-valued random variables the Kolmogorov-Rogozin inequality states that $$\mathcal{Q}_{h}(S_n)\leq C\left(\sum_{i=1}^{n}(1-\mathcal{Q}_{h}(X_i))\right)^{-\frac{1}{2}}.$$ In this paper we give an optimal bound for $\mathcal{Q}_{h}(S_n)$ in terms of $\mathcal{Q}_{h}(X_i)$, which settles a question posed by Leader and Radcliffe in 1994. Moreover, we show that the extremal distributions are mixtures of two uniform distributions each lying on an arithmetic progression.
Let $X_1,\ldots,X_n$ be independent identically distributed random vectors in $\mathbb{R}^d$. We consider upper bounds on $\max_x \mathbb{P}(a_1X_1+\cdots+a_nX_n=x)$ under various restrictions on $X_i$ and the weights $a_i$. When $\mathbb{P}(X_i=\pm 1) = \frac {1} {2}$, this corresponds to the classical Littlewood-Offord problem. We prove that in general for identically distributed random vectors and even values of $n$ the optimal choice for $(a_i)$ is $a_i=1$ for $i\leq \frac{n}{2}$ and $a_i=-1$ for $i > \frac {n} 2$, regardless of the distribution of $X_1$. Applying these results for Bernoulli random variables answers a recent question of Fox, Kwan and Sauermann. Finally, we provide sharp bounds for concentration probabilities of sums of random vectors under the condition $\sup_{x}\mathbb{P}(X_i=x)\leq α$, where it turns out that the worst case scenario is provided by distributions on an arithmetic progression that are in some sense as close to the uniform distribution as possible. An important feature of this work is that unlike much of the literature on the subject we use neither methods of harmonic analysis nor those from extremal combinatorics.
In a celebrated paper, Borwein, Erdélyi, Ferguson and Lockhart constructed cosine polynomials of the form \[ f_A(x) = \sum_{a \in A} \cos(ax), \] with $A\subseteq \mathbb{N}$, $|A|= n$ and as few as $n^{5/6+o(1)}$ zeros in $[0,2π]$, thereby disproving an old conjecture of J.E. Littlewood. Here we give a sharp analysis of their constructions and, as a result, prove that there exist examples with as few as $C(n\log n)^{2/3}$ roots.
Consider a sum $S_n=v_i\varepsilon_1+\cdots+v_n\varepsilon_{n}$, where $(v_i)^{n}_{i=1}$ are non-zero vectors in $\mathbb{R}^{d}$ and $(\varepsilon_i)^{n}_{i=1}$ are independent Rademacher random variables (i.e., $~{\mathbb{P}(\varepsilon_{i}=\pm 1)=1/2}$). The classical Littlewood-Offord problem asks for the best possible upper bound for $~{\sup_{x}\mathbb{P}(S_n = x)}$. In this paper we consider a non-uniform version of this problem. Namely, we obtain the optimal bound for $\mathbb{P}(S_n = x)$ in terms of the length of the vector $x\in \mathbb{R}^d$.
Majority bootstrap percolation on a graph $G$ is an epidemic process defined in the following manner. Firstly, an initially infected set of vertices is selected. Then step by step the vertices that have more infected than non-infected neighbours are infected. We say that percolation occurs if eventually all vertices in $G$ become infected. In this paper we study majority bootstrap percolation on the Erdős-Rényi random graph $G(n,p)$ above the connectivity threshold. Perhaps surprisingly, the results obtained for small $p$ are comparable to the results for the hypercube obtained by Balogh, Bollobás and Morris (2009).
Let S_n=X_1+...+X_n be a sum of independent symmetric random variables such that |X_{i}|\leq 1. Denote by W_n=ε_{1}+...+ε_{n} a sum of independent random variables such that \prob{\eps_i = \pm 1} = 1/2. We prove that \mathbb{P}{S_{n} \in A} \leq \mathbb{P}{cW_k \in A}, where A is either an interval of the form [x, \infty) or just a single point. The inequality is exact and the optimal values of c and k are given explicitly. It improves Kwapień's inequality in the case of the Rademacher series. We also provide a new and very short proof of the Littlewood-Offord problem without using Sperner's Theorem. Finally, an extension to odd Lipschitz functions is given.
Let $M_n= \fsu X1n$ be a sum of independent random variables such that $ X_k\leq 1$, $\E X_k =0$ and $\E X_k^2=\s_k^2$ for all $k$. Hoeffding 1963, Theorem 3, proved that $$¶{M_n \geq nt}\leq H^n(t,p),\quad H(t,p)= \bgl(1+qt/p\bgr)^{p +qt} \bgl({1-t}\bgr)^{q -qt}$$ with $$q=\ffrac 1{1+\s^2},\quad p=1-q, \quad \s^2 =\ffrac {\s_1^2+...+\s_n^2}n,\quad 0<t<1.$$ Bentkus 2004 improved Hoeffding's inequalities using binomial tails as upper bounds. Let $\ga_k =\E X_k^3/\s_k^3$ and $ \vk_k= \E X_k^4/\s_k^4$ stand for the skewness and kurtosis of $X_k$. In this paper we prove (improved) counterparts of the Hoeffding inequality replacing $\s^2$ by certain functions of $\fs \ga 1n$ respectively $\fs \vk 1n$. Our bounds extend to a general setting where $X_k$ are martingale differences, and they can combine the knowledge of skewness and/or kurtosis and/or variances of ~$X_k$. Up to factors bounded by $e^2/2$ the bounds are final. All our results are new since no inequalities incorporating skewness or kurtosis control so far are known.