Source author record

Sergey Dovgal

Explore this author’s publications, collaborators and research topics. This catalog record has not been claimed by its author.

ResearcherUnclaimed source record

Data Structures and Algorithms math.CO math.PR Discrete Mathematics math.ST Statistics Theory

Catalog footprint

Research overview

3works

6topics

4close collaborators

Actions

Is this your work?

Open graph Browse papers

preprint2022arXiv

The birth of the strong components

Random directed graphs $D(n,p)$ undergo a phase transition around the point $p = 1/n$ , and the width of the transition window has been known since the works of Luczak and Seierstad. They have established that as $n \to \infty$ when $p = (1 + μn^{-1/3})/n$ , the asymptotic probability that the strongly connected components of a random directed graph are only cycles and single vertices decreases from 1 to 0 as $μ$ goes from $-\infty$ to $\infty$ . By using techniques from analytic combinatorics, we establish the exact limiting value of this probability as a function of $μ$ and provide more properties of the structure of a random digraph around, below and above its transition point. We obtain the limiting probability that a random digraph is acyclic and the probability that it has one strongly connected complex component with a given difference between the number of edges and vertices (called excess). Our result can be extended to the case of several complex components with given excesses as well in the whole range of sparse digraphs. Our study is based on a general symbolic method which can deal with a great variety of possible digraph families, and a version of the saddle point method which can be systematically applied to the complex contour integrals appearing from the symbolic method. While the technically easiest model is the model of random multidigraphs, in which multiple edges are allowed, and where edge multiplicities are sampled independently according to a Poisson distribution with a fixed parameter $p$ , we also show how to systematically approach the family of simple digraphs, where multiple edges are forbidden, and where 2-cycles are either allowed or not. Our theoretical predictions are supported by numerical simulations, and we provide tables of numerical values for the integrals of Airy functions that appear in this study.

preprint2020arXiv

Counting directed acyclic and elementary digraphs

Directed acyclic graphs (DAGs) can be characterised as directed graphs whose strongly connected components are isolated vertices. Using this restriction on the strong components, we discover that when $m = cn$ , where $m$ is the number of directed edges, $n$ is the number of vertices, and $c < 1$ , the asymptotic probability that a random digraph is acyclic is an explicit function $p(c)$ , such that $p(0) = 1$ and $p(1) = 0$ . When $m = n(1 + μn^{-1/3})$ , the asymptotic behaviour changes, and the probability that a digraph is acyclic becomes $n^{-1/3} C(μ)$ , where $C(μ)$ is an explicit function of $μ$ . Łuczak and Seierstad (2009, Random Structures & Algorithms, 35(3), 271--293) showed that, as $μ\to -\infty$ , the strongly connected components of a random digraph with $n$ vertices and $m = n(1 + μn^{-1/3})$ directed edges are, with high probability, only isolated vertices and cycles. We call such digraphs elementary digraphs. We express the probability that a random digraph is elementary as a function of $μ$ . Those results are obtained using techniques from analytic combinatorics, developed in particular to study random graphs.

preprint2016arXiv

Towards Model Selection for Local Log-Density Estimation. Fisher and Wilks-type theorems

The aim of this research is to make a step towards providing a tool for model selection for log-density estimation. The author revisits the procedure for local log-density estimation suggested by Clive Loader (1996) and extends the theoretical results to finite-sample framework with the help of machinery of Spokoiny (2012). The results include bias expression from "deterministic" counterpart and Fisher and Wilks-type theorems from "stochastic". We elaborate on bandwidth trade-off $h(n) = \arg\min O(h^p) + O_p(1/\sqrt{nh^d})$ with explicit constants at big O notation. Explicit expressions involve (i) true density function and (ii) model that is selected (dimension, bandwidth, kernel and basis, e.g. polynomial). Existing asymptotic properties directly follow from our results. From the expressions obtained it is possible to control "the curse of dimension" both from the side of log-density smoothness and the inner space dimension.

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.