Plenary Lecture Prof. Robin C. Ball Saturday 12:30-13:30, MS02 Department of Physics and Centre for Complexity Science, University of Warwick Optimisation under Uncertainty It is a fact of life: we have to choose between complex options whose outcome is uncertain. I will focus on the following scenario: - the space of choices is too large to search exhaustively; - for each choice we do know the probability distribution of the outcome; - we have agreed what figure of merit we are trying to optimise, but we can only etimate this (with statistical errors!) by sampling from the distribution of outcomes. Three examples are how to design a protein molecule to spontaneously fold as fast as possible; how to plot a route of least expected length in advance of some stops becoming redundant (Probabilistic Travelling Salesman Problem); how most profitably to position the wells and pipelines to an oil field, before you know detail of where the oil will flow underground, the future price of oil, and installation costs. The problem is that the more accurately we evaluate each choice, the fewer choices we can explore. I will show how Simlulated Annealing by (Markov Chain) Monte Carlo can be adapted to deal with these problems. Simulated Annealing is directly analogous to cooling a physical system slowly, high merit corresponding to low energy in the final state. The key idea is that using rough estimates of the merit mimics an effective temperature: initially it is better to use rough estimates than more precise ones. The resulting technique has been termed Stochastic Annealing, and for known probability distribution of the errors the correspondence to true thermal equilibrium sampling can be made precise. This work was developed with (then!) research students Thomas Fink (now at Institut Curie, Paris) and Neil Bowler (now at UK MetOffice) [1]. Finally I will discuss a more open-ended example where we desire to optimise the worst likely case. The classic example is in Financial Portfolios, where this is expressed as the Value at Risk: in practice an arbitrary level of risk is adopted, say 1%, and we seek to minimise the Value there is 1% chance of losing. Stochastic Annealing can be used directly on this problem. However I will also discuss how an adaptation of it does something more intriguing: it optimises the value which appears to be at 0% risk on the basis of a limited depth of sampling of the distribution of outcomes. Thus instead of setting an arbitrary standard of risk, we set a limiting depth of search. This came out of work with Masters student Jonathan Mascie-Taylor [2]. [1] R.C. Ball, T.M.A. Fink, and N.E. Bowler, Stochastic Annealing, Physical Review Letters, 91,030201 (2003). [2] J. Mascie-Taylor, MSc Project, Centre for Complexity Science, University of Warwick (2009). Keynote Lectures Combinatorics Prof. Peter Cameron Thursday 14:00-15:00, MS03 School of Mathematical Sciences, Queen Mary, University of London Synchronization The topic of synchronization arose in automata theory, but has connections with permutation groups and graph homomorphisms. I will discuss some of these, and also an application of Joyal's proof of Cayley's Theorem on trees to the synchronization behaviour of a random single-transition automaton. Mathematical Biology Prof. Raymond Goldstein Thursday 14:00-15:00, MS04 DAMTP, University of Cambridge Fluid Dynamics and the Evolution of Biological Complexity One of the most fundamental issues in evolutionary biology is the nature of transitions from unicellular organisms to multicellular ones. Many basic questions arise in this context: What is the advantage of being larger? What are the driving forces behind the appearance of distinct cell types? In this talk I will describe an approach to these broad questions based on the use of a particular lineage of green algae which serves not only as a model for evolutionary studies, but also for biological fluid dynamics. Experimental and theoretical results will be described that focus on the scaling laws for nutrient uptake as a function of size, particularly as affected by fluid flows driven by the flagella of these organisms, the synchronization of those flagella, and the mechanism by which multicellular organisms composed of thousands of cells exhibit accurate phototaxis in the absence of a central nervous system. Mathematical Physics Prof. Sandra Chapman, FInstP Thursday 14:00-15:00, MS02 Physics Department, University of Warwick Scaling laws, emergence and statistical descriptions of systems that are out of equilibrium: what we can model and measure Physics aims to find descriptions of natural systems that are universal. From the starting point of dimensional analysis, through simple CA models that show emergence, to what can be observed in data, we will explore out of equilibrium systems with examples including astrophysical turbulence, laboratory plasmas, ecology and neuroscience. We will pose the question- what are the prospects for universal theories of out of equilibrium systems that can be tested on real world data? Differential Geometry Prof. Nigel Hitchin, FRS Thursday 14:00-15:00, MS03 Mathematical Institute, University of Oxford Generalized geometry and Poisson geometry Generalized geometry is a differential geometric structure which seems to contain naturally some of the features of supersymmetric theories but can be viewed in very concrete terms and in particular one can draw on analogies with ordinary complex or Riemannian geometry. On the other hand the role of Poisson structures, especially holomorphic ones, also turns out to be an important aspect. The talk will discuss the interrelationship between the two, and how ideas from one can suggest approaches in the other. Complexity Prof. Jeff Johnson Thursday 17:30-18:30, MS03 Open University Hypernetworks for modelling multilevel complex systems Hypergraphs generalise graphs by allowing edges to be sets with any number of of vertices. Although a big step forward, hypergraphs are set-theoretic constructs with anomalies such as {D, O, G} = {G, O, D}. Hypernetworks overcome this by the definition of 'relational simplices', <a, b, c, ...; R> where the n vertices a, b, c, ... are related under the n-ary relation R. For example, <mum, dad, son, daughter; R_family> forms a social construct called a family. Here the object with four vertices is represented as a 3-dimensional tetrahedron, a higher dimensional analogue to the 1-dimensional edges in networks. This example also illustrates how n-ary relations can define different levels of aggregation, with the parts being at lower levels to wholes. Relational simplices have a higher dimensional connectivity structure, for example two tetrahedra sharing a triangular face are 2-connected, a higher dimensional analogue to edges being 0-dimensionally connected by vertices in networks. Simplices carry patterns of numbers over their vertices and faces that can represent dynamics such as flows or system activity. These dynamics are different to the structural dynamics of the formation and disintegration of relational simplices. In this talk I will show how hypernetworks of relational simplices can represent relational structure and dynamics in complex multilevel systems. The emphasis will be on ideas rather than technical details, and illustrations will be given from social, physical and engineering systems. Algebra Prof. Martin R. Bridson Friday 15:00-16:00, MS05 Mathematical Institute, University of Oxford The universe of finitely presented groups Symmetry is everywhere in mathematics and the language in which to describe and examine symmetry is that of group theory. How should one describe (infinite, discrete) groups? How should one go about understanding them? Can one ultimately hope to understand all (finitely presented) groups completely, or will undecidability phenomena necessarily intrude at some point? Can one use geometry to understand an arbitrary group, and if so, what geometry? Can one hope to understand infinite groups by understanding all of their finite images? Lots of questions! In this talk, I'll describe the broad outlines of the modern study of infinite discrete groups, sketch the universe of finitely presented groups, present answers to some of the preceding questions, and point to some of the most exciting directions of current research. Financial Mathematics Prof. David Hand, FBA Friday 15:00-16:00, MS04 Department of Mathematics, Imperial College Evaluating consumer credit scoring models Predictive statistical models are universally used in the consumer credit industry to guide decisions about who should be offered financial products such as loans, credit cards, mortgages, car finance, and so on, and also to monitor the behaviour of customers to see if they are running into difficulties and to detect fraud. In this talk I examine such models, with a particular emphasis on their evaluation: accurate evaluation is critical to ensure that correct commercial decisions are made. Some recent results are presented. Statistics and Probability Prof. Saul Jacka Friday 15:00-16:00, MS02 Department of Statistics, University of Warwick Stochastic Control and Applications Stochastic Control is a slightly neglected weapon in the probabilist's (and occasionally the statistician's) armoury. The talk will focus on some basics of stochastic control and some representative applications. Topology Prof. Caroline Series Friday 15:00-16:00, MS03 Warwick Mathematics Institute, University of Warwick Limits of limit sets A Kleinian group is a discrete group of isometries of hyperbolic 3-space. Its limit set is the set of points where orbits accumulate on the boundary of hyperbolic space, which we identify with the Riemann sphere. As the group varies continuously, we explore how limit sets can degenerate, even collapsing to a space filling curve. Is it true that limit sets always move continuously with the group? Algebraic Geometry Prof. Burt Totaro, FRS Saturday 10:15-11:15, MS02 DPMMS, University of Cambridge Algebraic geometry from a topological point of view We discuss the role of topology in algebraic geometry. There are simple but surprising arguments which show that many properties of a complex submanifold of projective space are controlled by its topology, in fact by the second homology group. More subtle properties are determined by a convex cone, the "cone of curves". For algebraic surfaces, the cone of curves can be studied using hyperbolic geometry. Number Theory Prof. Kevin Buzzard Saturday 11:30-12:30, MS02 Department of Mathematics, Imperial College Future directions in the Langlands program. The Langlands program is a profound series of conjectures about analytic gadgets called automorphic forms. Some of these conjectures relate these analytic gadgets to much more algebraic gadgets (elliptic curves, Galois groups and so on). More recently there has been some serious attempt to try and formulate p-adic versions of these conjectures, which should in some sense be easier because p-adic analysis is far more "algebraic" than real analysis. I will attempt to give some sort of an overview of both the classical, and the more modern p-adic, theories. Be warned: I have to start somewhere, and I don't want to spend 30 minutes on basic definitions and properties of the p-adic numbers, so I will spend under 5 minutes on such things and then press on. On the other hand I will give precise definitions of things like automorphic forms and not expect the audience to be familiar with Langlands program at all. Analysis and PDEs Sir John Ball, FRS Saturday 11:30-12:30, MS03 Mathematical Institute, University of Oxford Variational problems for solid and liquid crystals The talk will describe how the calculus of variations can provide insights into the behaviour of solid and liquid crystals, focussing on some open problems. Dynamical Systems Prof. Mary Rees, FRS Saturday 11:30-12:30, MS04 Department of Mathematical Sciences, Liverpool University Topological models in complex dynamics A general problem in complex dynamics is to describe a dynamical system up to change of coordinates, that is, up to (at least topological) conjugacy. This is an especially big topic in complex dynamics, when a large variety of examples is easy to exhibit in a single parameter space. I shall describe some of the common constructions, and some of the problems that arise.