Past Combinatorial Theory Seminar

We consider two decision problems for linear recurrence sequences(LRS) over the integers, namely the Positivity Problem (are all terms of a given LRS positive?) and the Ultimate Positivity Problem (are all but finitely many terms of a given LRS positive?). We show decidability of both problems for LRS of order 5 or less, and for simple LRS (i.e. whose characteristic polynomial has no repeated roots) of order 9 or less. Moreover, we show by way of hardness that extending the decidability of either problem to LRS of order 6 would entail major breakthroughs in analytic number theory, more precisely in the field of Diophantine approximation of transcendental numbers.
This talk is based on a recent paper, available at
http://www.cs.ox.ac.uk/people/joel.ouaknine/publications/positivity13ab…
joint with James Worrell and Matt Daws.

The recently completed auction for 4G mobile spectrum was the most importantcombinatorial auction ever held in the UK.  In general, combinatorial auctions allow bidders to place individual bids on packages of items,instead of separate bids on individual items, and this feature has theoretical advantages for bidders and sellers alike.  The accompanying challenges of implementation have been the subject of intense work over the last few years, with the result that the advantages of combinatorial auctions can now be realised in practice on a large scale.  Nowhere has this work been more prominent than in auctions for radio spectrum.  The UK's 4G auction is the most recent of these and the publication by Ofcom (the UK's telecommunications regulator) of the auction's full bidding activity creates a valuable case study of combinatorial auctions in action.

We prove that if $\frac{\log^{117} n}{n} \leq p \leq 1 -

n^{-1/8}$, then asymptotically almost surely the edges of $G(n,p)$ can

be covered by $\lceil \Delta(G(n,p))/2 \rceil$ Hamilton cycles. This

is clearly best possible and improves an approximate result of Glebov,

Krivelevich and Szab\'o, which holds for $p \geq n^{-1 + \varepsilon}$.

Based on joint work with Daniela Kuhn, John Lapinskas and Deryk Osthus.

Razborov's flag algebras provide a formal system

for operating with asymptotic inequalities between subgraph densities,

allowing to do extensive "book-keeping" by a computer. This novel use

of computers led to progress on many old problems of extremal

combinatorics. In some cases, finer structural information can be

derived from a flag algebra proof by by using the Removal Lemma or

graph limits. This talk will overview this approach.

While usual percolation concerns the study of the connected components of

random subgraphs of an infinite graph, bootstrap percolation is a type of

cellular automaton, acting on the vertices of a graph which are in one of

two states: `healthy' or `infected'. For any positive integer $r$, the

$r$-neighbour bootstrap process is the following update rule for the

states of vertices: infected vertices remain infected forever and each

healthy vertex with at least $r$ infected neighbours becomes itself

infected. These updates occur simultaneously and are repeated at discrete

time intervals. Percolation is said to occur if all vertices are

eventually infected.

As it is often difficult to determine precisely which configurations of

initially infected vertices percolate, one often considers a random case,

with each vertex infected independently with a fixed probability $p$. For

an infinite graph, of interest are the values of $p$ for which the

probability of percolation is positive. I will give some of the history

of this problem for regular trees and present some new results for

bootstrap percolation on certain classes of randomly generated trees:

Galton--Watson trees.

I will describe how a search for the answer to an old question about the existence of monotone arithmetic progressions in permutations of positive integers led to the study of infinite words with bounded additive complexity. The additive complexity of a word on a finite subset of integers is defined as the function that, for a positive integer $n$, counts the maximum number of factors of length $n$, no two of which have the same sum.

Results of Bourgain and Kindler-Safra state that if $f$ is a Boolean function on $\{0,1\}^n$, and

the Fourier transform of $f$ is highly concentrated on low frequencies, then $f$ must be close

to a ‘junta’ (a function depending upon a small number of coordinates). This phenomenon is

known as ‘Fourier stability’, and has several interesting consequences in combinatorics,

theoretical computer science and social choice theory. We will describe some of these,

before turning to the analogous question for Boolean functions on the symmetric group. Here,

genuine stability does not occur; it is replaced by a weaker phenomenon, which we call

‘quasi-stability’. We use our 'quasi-stability' result to prove an isoperimetric inequality

for $S_n$ which is sharp for sets of size $(n-t)!$, when $n$ is large. Several open questions

remain. Joint work with Yuval Filmus (University of Toronto) and Ehud Friedgut (Weizmann

Institute).

A self-avoiding walk on a lattice is a walk that never visits the same vertex twice.  Self-avoiding walks (SAW) have attracted interest for decades, first in statistical physics, where they are considered as polymer models, and then in combinatorics and in probability theory (the first mathematical contributions are probably due to John Hammersley, from Oxford, in the early sixties). However, their properties remain poorly understood in low dimension, despite the existence of remarkable conjectures.

About two years ago, Duminil-Copin and Smirnov proved an "old" and remarkable conjecture of Nienhuis (1982), according to which the number of SAWs of length n on the honeycomb (hexagonal) lattice grows like mu^n, with mu=sqrt(2 +sqrt(2)).

This beautiful result has woken up the hope to prove other simple looking conjectures involving these objects. I will thus present the proof of a younger conjecture (1995) by Batchelor and Yung, which deals with SAWs confined to a half-plane and interacting with its boundary.

(joint work with N. Beaton, J. de Gier, H. Duminil-Copin and A. Guttmann)

Let $Q_n$ denote the graph of the $n$-dimensional cube with vertex set $\{0, 1\}^n$

in which two vertices are adjacent if they differ in exactly one coordinate.

Suppose $G$ is a subgraph of $Q_n$ with average degree at least $d$. How long a

path can we guarantee to find in $G$?

My aim in this talk is to show that $G$ must contain an exponentially long

path. In fact, if $G$ has minimum degree at least $d$ then $G$ must contain a path

of length $2^d − 1$. Note that this bound is tight, as shown by a $d$-dimensional

subcube of $Q^n$. I hope to give an overview of the proof of this result and to

discuss some generalisations.

Rainbow connectivity is a new concept for measuring the connectivity of a graph which was introduced in 2008 by Chartrand, Johns, McKeon and Zhang. In a graph G with a given edge colouring, a rainbow path is a path all of whose edges have distinct colours. The minimum number of colours required to colour the edges of G so that every pair of vertices is joined by at least one rainbow path is called the rainbow connection number rc(G) of the graph G.

For any graph G, rc(G) >= diam(G). We will discuss rainbow connectivity in the random graph setting and present the result that for random graphs, rainbow connectivity 2 happens essentially at the same time as diameter 2. In fact, in the random graph process, with high probability the hitting times of diameter 2 and of rainbow connection number 2 coincide

Given a finite list of vectors X in $\R^d$, one can define the box spline $B_X$. Box splines are piecewise polynomial functions that are used in approximation theory. They are also interesting from a combinatorial point of view and many of their properties solely depend on the structure of the matroid defined by the list X. The support of the box spline is a certain polytope called zonotope Z(X). We will show that if the list X is totally unimodular, any real-valued function defined on the set of lattice points in the interior of Z(X) can be extended to a function on Z(X) of the form $p(D)B_X$ in a unique way, where p(D) is a differential operator that is contained in the so-called internal P-space. This was conjectured by Olga Holtz and Amos Ron. The talk will focus on combinatorial aspects and all objects mentioned above will be defined. (arXiv:1211.1187)

In this talk we present a general method using permanent estimates in order to obtain results about counting and packing Hamilton cycles in dense graphs and oriented graphs. As a warm up we prove that every Dirac graph $G$ contains at least $(reg(G)/e)^n$ many distinct Hamilton cycles, where $reg(G)$ is the maximal degree of a spanning regular subgraph of $G$. We continue with strengthening a result of Cuckler by proving that the number of oriented Hamilton cycles in an almost $cn$-regular oriented graph is $(cn/e)^n(1+o(1))^n$, provided that $c$ is greater than $3/8$. Last, we prove that every graph $G$ of minimum degree at least $n/2+\epsilon n$ contains at least $reg_{even}(G)-\epsilon n$ edge-disjoint Hamilton cycles, where $reg_{even}(G)$ is the maximal even degree of a spanning regular subgraph of $G$. This proves an approximate version of a conjecture made by Osthus and K\"uhn.  Joint work with Michael Krivelevich and Benny Sudakov.

In an Erdős--R\'enyi random graph above the phase transition, i.e.,

where there is a giant component, the size of (number of vertices in)

this giant component is asymptotically normally distributed, in that

its centred and scaled size converges to a normal distribution. This

statement does not tell us much about the probability of the giant

component having exactly a certain size. In joint work with B\'ela

Bollob\'as we prove a `local limit theorem' answering this question

for hypergraphs; the graph case was settled by Luczak and Łuczak.

The proof is based on a `smoothing' technique, deducing the local

limit result from the (much easier) `global' central limit theorem.

Results that say local information is enough to guarantee global information provide the theoretical underpinnings of many reconstruction algorithms in evolutionary biology. Such results include Buneman's Splits-Equivalence Theorem and the Tree-Metric Theorem. The first result says that, for a collection $\mathcal C$ of binary characters, pairwise compatibility is enough to guarantee compatibility for $\mathcal C$, that is, there is a phylogenetic (evolutionary) tree that realises $\mathcal C$. The second result says that, for a distance matrix $D$, if every $4\times 4$ distance submatrix of $D$ is realisable by an edge-weighted phylogenetic tree, then $D$ itself is realisable by such a tree. In this talk, we investigate these and other results of this type. Furthermore, we explore the closely-related task of determining how much information is enough to reconstruct the correct phylogenetic tree.

We study the problem of covering R^d by overlapping translates of a convex polytope, such that almost every point of R^d is covered exactly k times. Such a covering of Euclidean space by a discrete set of translations is called a k-tiling. The investigation of simple tilings by translations (which we call 1-tilings in this context) began with the work of Fedorov and Minkowski, and was later extended by Venkov and McMullen to give a complete characterization of all convex objects that 1-tile R^d. By contrast, for k ≥ 2, the collection of polytopes that k-tile is much wider than the collection of polytopes that 1-tile, and there is currently no known analogous characterization for the polytopes that k-tile. Here we first give the necessary conditions for polytopes P that k-tile, by proving that if P k-tiles R^d by translations, then it is centrally symmetric, and its facets are also centrally symmetric. These are the analogues of Minkowski’s conditions for 1-tiling polytopes, but it turns out that very new methods are necessary for the development of the theory. In the case that P has rational vertices, we also prove that the converse is true; that is, if P is a rational, centrally symmetric polytope, and if P has centrally symmetric facets, then P must k-tile R^d for some positive integer k.

We call a graph $H$ \emph{Ramsey-unsaturated} if there is an edge in the

complement of $H$ such that the Ramsey number $r(H)$ of $H$ does not

change upon adding it to $H$. This notion was introduced by Balister,

Lehel and Schelp who also showed that cycles (except for $C_4$) are

Ramsey-unsaturated, and conjectured that, moreover, one may add {\em

any} chord without changing the Ramsey number of the cycle $C_n$, unless

$n$ is even and adding the chord creates an odd cycle.

We prove this conjecture for large cycles by showing a stronger

statement: If a graph $H$ is obtained by adding a linear number of

chords to a cycle $C_n$, then $r(H)=r(C_n)$, as long as the maximum

degree of $H$ is bounded, $H$ is either bipartite (for even $n$) or

almost bipartite (for odd $n$), and $n$ is large.

This motivates us to call cycles \emph{strongly} Ramsey-unsaturated.

Our proof uses the regularity method.

Graphs and digraphs behave quite differently, and many classical results for graphs are often trivially false when extended to general digraphs. Therefore it is usually necessary to restrict to a smaller family of digraphs to obtain meaningful results. One such very natural family is Eulerian digraphs, in which the in-degree equals out-degree at every vertex.

In this talk, we discuss several natural parameters for Eulerian digraphs and study their connections. In particular, we show that for any Eulerian digraph G with n vertices and m arcs, the minimum feedback arc set (the smallest set of arcs whose removal makes G acyclic) has size at least $m^2/2n^2+m/2n$, and this bound is tight. Using this result, we show how to find subgraphs of high minimum degrees, and also long cycles in Eulerian digraphs. These results were motivated by a conjecture of Bollob\'as and Scott.

Joint work with Ma, Shapira, Sudakov and Yuster

It is very well known that every graph on $n$ vertices and $m$ edges admits a bipartition of size at least $m/2$. This bound can be improved to $m/2 + (n-1)/4$ for connected graphs, and $m/2 + n/6$ for graphs without isolated vertices, as proved by Edwards, and Erd\"os, Gy\'arf\'as, and Kohayakawa, respectively. A bisection of a graph is a bipartition in which the size of the two parts differ by at most 1. We prove that graphs with maximum degree $o(n)$ in fact admit a bisection which asymptotically achieves the above bounds.These results follow from a more general theorem, which can also be used to answer several questions and conjectures of Bollob\'as and Scott on judicious bisections of graphs.
Joint work with Po-Shen Loh and Benny Sudakov

Random geometric graphs have been well studied over the last 50 years or so. These are graphs that

are formed between points randomly allocated on a Euclidean space and any two of them are joined if

they are close enough. However, all this theory has been developed when the underlying space is

equipped with the Euclidean metric. But, what if the underlying space is curved?

The aim of this talk is to initiate the study of such random graphs and lead to the development of

their theory. Our focus will be on the case where the underlying space is a hyperbolic space. We

will discuss some typical structural features of these random graphs as well as some applications,

related to their potential as a model for networks that emerge in social life or in biological

sciences.

We discuss some recent developments on the following long-standing problem known as Ryser's

conjecture. Let $H$ be an $r$-partite $r$-uniform hypergraph. A matching in $H$ is a set of disjoint

edges, and we denote by $\nu(H)$ the maximum size of a matching in $H$. A cover of $H$ is a set of

vertices that intersects every edge of $H$. It is clear that there exists a cover of $H$ of size at

most $r\nu(H)$, but it is conjectured that there is always a cover of size at most $(r-1)\nu(H)$.

Rado introduced the following `lion and man' game in the 1930's: two players (the lion and the man) are in the closed unit disc and they can run at the same speed. The lion would like to catch the man and the man would like to avoid being captured.

This game has a chequered history with several false `winning strategies' before Besicovitch finally gave a genuine winning strategy.

We ask the surprising question: can both players win?

A dot product representation of a graph assigns to each vertex $s$ a vector $v(s)$ in ${\bf R}^k$ in such a way that $v(s)^T v(t)$ is greater than $1$ if and only $st$ is an edge. Similarly, in a distance representation $|v(s)-v(t)|$ is less than $1$ if and only if $st$ is an edge.

I will discuss the solution of some open problems by Spinrad, Breu and Kirkpatrick and others on these and related geometric representations of graphs. The proofs make use of a connection to oriented pseudoline arrangements.

(Joint work with Colin McDiarmid and Ross Kang)

If $A$ is a set of $n$ positive integers, how small can the set

$\{ x/(x,y) : x,y \in A \}$ be? Here, as usual, $(x,y)$ denotes the highest common factor of

$x$ and $y$. This elegant question was raised by Granville and Roesler, who

also reformulated it in the following way: given a set $A$ of $n$ points in

the integer grid ${\bf Z}^d$, how small can $(A-A)^+$, the projection of the difference

set of $A$ onto the positive orthant, be?

Freiman and Lev gave an example to show that (in any dimension) the size can

be as small as $n^{2/3}$ (up to a constant factor). Granville and Roesler

proved that in two dimensions this bound is correct, i.e. that the size is

always at least $n^{2/3}$, and they asked if this holds in any dimension.

After some background material, the talk will focus on recent developments.

Joint work with B\'ela Bollob\'as.

In Achlioptas processes, starting from an empty graph, in each step two potential edges are chosen uniformly at random, and using some rule one of them is selected and added to the evolving graph. Although the evolution of such `local' modifications of the Erdös-Rényi random graph processes has received considerable attention during the last decade, so far only rather `simple' rules are well-understood. Indeed, the main focus has been on bounded size rules (where all component sizes larger than some constant B are treated the same way), and for more complex rules hardly any rigorous results are known. In this talk we will discuss a new approach that applies to many involved Achlioptas processes: it allows us to prove that certain key statistics are tightly concentrated during the early evolution of e.g. the sum and product rule.

Joint work with Oliver Riordan.