36 DM14 Abstracts IP0 University of Oxford Hot Topic Session: The Existence of Designs [email protected] We prove the existence conjecture for combinatorial de- signs, answering a question of Steiner from 1853. More IP4 generally, we show that the natural divisibility conditions Submodular Functions and Their Applications are sufficient for clique decompositions of simplicial com- plexes that satisfy a certain pseudorandomness condition. Submodular functions, a discrete analogue of convex func- tions, have played a fundamental role in combinatorial op- timization since the 1970s. In the last decade, there has Peter Keevash been renewed interest in submodular functions due to their University of Oxford role in algorithmic game theory, as well as numerous appli- [email protected] cations in machine learning. These developments have led to new questions as well as new algorithmic techniques. I will discuss the concept of submodularity, its unifying role IP1 in combinatorial optimization, and a few illustrative appli- The Graph Regularity Method cations. I will describe some recent algorithmic advances, in particular the concept of multilinear relaxation, and the Szemerdi’s regularity lemma is one of the most powerful role of symmetry in proving hardness results for submod- tools in graph theory, with many applications in combina- ular optimization. I will conclude by discussing possible torics, number theory, discrete geometry, and theoretical extensions of the notion of submodularity, and some fu- computer science. Roughly speaking, it says that every ture challenges. large graph can be partitioned into a small number of parts such that the bipartite subgraph between almost all pairs Jan Vondrak of parts is random-like. Several variants of the regularity IBM Almaden Research Center lemma have since been established with many further ap- [email protected] plications. I will discuss recent progress in understanding the quantitative aspects of these lemmas and their appli- cations. IP5 Approximation Algorithms Via Matrix Covers Jacob Fox MIT I will describe a general technique for obtaining approx- [email protected] imate solutions of hard quadratic optimization problems using economical covers of high dimensional sets by small cubes. The analysis of the method leads to intriguing al- IP2 gorithmic, combinatorial, geometric, extremal and proba- Rota’s Conjecture bilistic questions. Based on joint papers with Troy Lee, Adi Shraibman and Santosh Vempala. In 1970, Gian-Carlo Rota posed a conjecture giving a suc- cinct combinatorial characterization of the linear depen- Noga Alon dencies among a finite set of vectors in a vector space over Tel-Aviv University any given finite field. I will discuss the conjecture as well [email protected] as joint work with Bert Gerards and Geoff Whittle that led to a solution. IP6 Jim Geelen Interlacing Families and Kadison–Singer University of Waterloo [email protected] This talk will focus on the resolution of two conjectures that have been shown to be equivalent to open problems in a number of fields, most famously a problem due to Kadi- IP3 son and Singer concerning the foundations of mathematical Covering Systems of Congruences physics. This will include introducing a new technique for establishing the existence of certain combinatorial objects A distinct covering system of congruences is a collection that we call the ”method of interlacing polynomials.” The ai mod mi,1<m1 <m2 < ... < mk, whose union is the technique seems to be interesting in its own right (it was integers. Covering systems were introduced by Paul Erdos, also the main tool for resolving the existence of Ramanujan who used the system graphs of arbitrary degree). This represents join work with Z =(0mod2)∪(0 mod 3)∪(1 mod 4)∪(3 mod 8)∪(7 mod 12)∪Dan(23 Spielmanmod 24) of Yale University and Nikhil Srivastava of Microsoft Research, India. to produce an odd arithmetic progression none of whose members is of the form a prime plus a power of two. Two Adam Marcus well-known questions of Erdos concern covering systems. Yale University The minimum modulus problem asks if there exist distinct [email protected] covering systems for which m1 is arbitrarily large. The odd modulus problem asks for a distinct covering system with all moduli odd. I will describe aspects of my negative an- IP7 swer to the minimum modulus problem and ongoing joint Hamilton Decompositions of Graphs and Digraphs work with Pace Nielsen toward the odd modulus problem. The arguments involve techniques from graph theory and In this talk I will survey recent results on Hamilton decom- PDE. positions, i.e. decompositions of the edge-set of a graph or digraph into edge-disjoint Hamilton cycles. One example Robert Hough is a conjecture of Kelly from 1968, which states that every DM14 Abstracts 37 regular tournament has a Hamilton decomposition. We Jozsef Balogh prove this conjecture for large tournaments. In fact, we University of Illinois prove a far more general result, which has further appli- [email protected] cations. Another example is the Hamilton decomposition conjecture, which was posed by Nash-Williams in 1970. It states that every regular graph on n vertices with mini- CP1 mum degree at least n/2 has a Hamilton decomposition Percolation in the Weighted Random Connection (joint work with Bela Csaba, Allan Lo, Deryk Osthus and Model Andrew Treglown). We propose the weighted random connection model, with Daniela Kuhn the goal to predict and better understand the spread of in- University of Birmingham fectious diseases. Each individual of a population is charac- [email protected] terized by two parameters: its position and risk behavior, where the probability of transmissions among individuals increases in the individual risk factors, and decays in their IP8 Euclidean distance. Our main result is on the almost sure The Complexity of Graph and Hypergraph Expan- existence of an infinite component in the weighed random sion Problems connection model, which corresponds to the outbreak of an infectious disease. We use results on the random connec- The expansion of a graph is a fundamental parameter with tion model and site percolation in Z2. many connections in mathematics and applications in com- puter science. Two basic variants of expansion are edge- Milan Bradonjic expansion (minimum ratio of the number of edges leav- Mathematics of Networks and Communications ing a subset of vertices to the total number of edge inci- Bell Labs, Alcatel-Lucent dent to the subset) and vertex expansion (minimum ratio [email protected] of number of vertices adjacent to a subset to the size of the subset). Both are NP-hard to compute exactly and CP1 the current√ best polynomial-time algorithms provide ei- ther an O( log n)OPT√ solution these problems, or, for Degeneracy in R-Mat Generated Graphs edge-expansion, an O( OPT ) solution. The latter is via Cheeger’s inequality relating edge expansion to the second The degeneracy of a graph G is the minimum possible value smallest eigenvalue of the graph Laplacian, and provides of d such that every subgraph of G has a vertex with de- an efficient method to check if a graph is an edge expander gree at most d. It is known that the degeneracy of an (i.e., has edge expansion is at least some constant). In this Erd˝os-R´enyi random graph is bounded with high probabil- talk, we survey recent developments on the following lines: ity. We examine degeneracy in the popular recursive ma- 1. What is the analog of expansion for partitioning a graph trix (R-MAT) random graph model, of which Erd˝os-R´enyi into multiple parts (more than 2), and how does Cheeger’s is a special case, and determine whether its asymptotic inequality generalize? 2. Is there a Cheeger-type inequal- behavior parallels that of Erd˝os-R´enyi graphs. ity for vertex expansion? How to efficiently verify whether Alex Chin a graph is a vertex expander? 3. Do there exist algorithms North Carolina State University with better performance? 4. How do these parameters, [email protected] inequalities and algorithms extend to hypergraphs? Santosh Vempala Timothy Goodrich Georgia Institute of Technology Valparaiso University [email protected] [email protected] Blair Sullivan SP1 North Carolina State University 2014 Dnes Knig Prize Lecture-The number of Ks,t- [email protected] free graphs The problem of enumerating H-free graphs with a given CP1 number of vertices has been studied for almost forty years. Identifying Codes and Searching with Balls in As a result, a great deal is known about this problem. Graphs When H is not bipartite, we have not only good estimates of the number of H-free graphs, but also fairly precise We address the following search theory problem: what is structural characterisation of a typical such graph. On the the minimum number of queries that is needed to deter- contrary, not much is known in case when H is bipartite. mine an unknown vertex in a graph G if queries are balls A few years ago, we proved asymptotically optimal upper with respect to the graph distance and the answer to a bounds on the number of graphs not containing a fixed query is YES if the unknown vertex lies in the ball and complete bipartite graph Ks,t. This was the first infinite NO otherwise. We study adaptive or non-adaptive versions family of bipartite graphs for which such a bound was es- of this problem for the following classes of graphs: hyper- tablished. Since then, the method developed in our proof cubes, the Erd˝os-R´enyi random graph model and graphs has found several interesting applications to a large class of bounded maximum degree. of related problems.