Past

The applied, numerical and asymptotic analysis of acoustic, electromagnetic and elastic scattering by smooth scatterers (e.g. a cylinder or a sphere) is a classical topic in applied mathematics. However, many real-world applications involve highly non-smooth scatterers with geometric structure on multiple length scales. Examples include acoustic scattering by trees and other vegetation in the modelling of urban noise propagation, electromagnetic scattering by snowflakes and ice crystal aggregates in climate modelling and weather prediction, and elastic scattering by cracks and other interfaces in seismic imaging and hydrocarbon exploration. In such situations it may be more appropriate to model the scatterer not by a smooth surface but by a fractal, a geometric object with self-similarity properties and detail on every length scale. Well-known examples include the Cantor set, Sierpinski triangle and the Koch snowflake. In this talk I will give an overview of our recent research into acoustic scattering by such fractal structures. So far our work has focussed on establishing well-posedness of the scattering problem and integral equation reformulations of it, and developing and analysing numerical methods for obtaining approximate solutions. However, there remain interesting open questions about the high frequency (short wavelength) asymptotic behaviour of solutions, and whether the self-similarity of the scatterer can be exploited to derive more efficient approximation techniques.

Separability is an algebraic property enjoyed by certain subsets of groups. In the world of non-positively curved groups, it has a not-too-well-understood link to geometric properties such as convexity. We explore this connection in the setting of relatively hyperbolic groups and discuss a recent joint work in this area involving products of quasiconvex subgroups.

In this talk, I will present a noncommutative analogue of Margulis’ factor theorem for higher rank lattices. More precisely, I will give a complete description of all intermediate von Neumann subalgebras sitting between the von Neumann algebra of the lattice and the von Neumann algebra of the action of the lattice on the Furstenberg-Poisson boundary. As an application, we infer that the rank of the semisimple Lie group is an invariant of the pair of von Neumann algebras. I will explain the relevance of this result regarding Connes’ rigidity conjecture.

I will present recent results on the growth of entanglement between two adjacent regions in a tripartite, one-dimensional many-body system after a quantum quench. Combining a replica trick with a space-time duality transformation a universal relation between the entanglement negativity and Renyi-1/2 mutual information can be derived, which holds at times shorter than the sizes of all subsystems. The proof is directly applicable to any local quantum circuit, i.e., any lattice system in discrete time characterised by local interactions, irrespective of the nature of its dynamics. The derivation indicates that such a relation can be directly extended to any system where information spreads with a finite maximal velocity. The talk is based on Phys. Rev. Lett. 129, 140503 (2022).

Various graphs associated to surfaces have proved to be important tools for studying the large scale geometry of mapping class groups of surfaces, among other applications. A seminal paper of Masur and Minsky proved that perhaps the most well known example, the curve graph, is Gromov hyperbolic. However, this is not the case for every naturally defined graph associated to a surface. We will present joint work with Jacob Russell classifying a wide family of graphs associated to surfaces according to whether the graph is Gromov hyperbolic, relatively hyperbolic or not relatively hyperbolic.
 

A finite W-algebra is a gadget associated to each nilpotent orbit in a complex semisimple Lie algebra g. There is a functor from W-modules to a full subcategory of g-modules, known as Skryabin’s equivalence, and every primitive ideals of the enveloping algebra U(g) as the annihilator of a module obtained in this way. This gives a convenient way of organising together primitive ideals in terms of nilpotent orbits, and this approach has led to a resurgence of interest in some hard open problems which lay dormant for some 20 years. The primitive ideals of U(g) which come from one-dimensional representations of W-algebras are especially nice, and we shall call them Losev—Premet ideals. The goal of this talk is to explain my recent work which seeks to: (1) describe the structure of the space of the dimensional representations of a finite W-algebra and (2) classify the Losev—Premet ideals.

A random $d$-regular graph is just a $d$-regular simple graph on $[n]=\{1,2,\ldots,n\}$ chosen uniformly at random from all such graphs. This model, with $d=d(n)$, is one of the most natural random graph models, but is quite tricky to work with/reason about, since actually generating such a graph is not so easy. For $d$ constant, Bollobás's configuration model works well; for larger $d$ one can combine this with switching arguments pioneered by McKay and Wormald. I will discuss recent progress with Nick Wormald, pushing linear-time generation up to $d=o(\sqrt{n})$. One ingredient is reciprocal rejection sampling, a trick for 'accepting' a certain graph with a probability proportional to $1/N(G)$, where $N(G)$ is the number of certain configurations in $G$. The trick allows us to do this without calculating $N(G)$, which would take too long.

Many networks exhibit some community structure. There exists a wide variety of approaches to detect communities in networks, each offering different interpretations and associated algorithms. For large networks, there is the additional requirement of speed. In this context, the so-called label propagation algorithm (LPA) was proposed, which runs in near linear time. In partitions uncovered by LPA, each node is ensured to have most links to its assigned community. We here propose a fast variant of LPA (FLPA) that is based on processing a queue of nodes whose neighbourhood recently changed. We test FLPA exhaustively on benchmark networks and empirical networks, finding that it runs up to 700 times faster than LPA. In partitions found by FLPA, we prove that each node is again guaranteed to have most links to its assigned community. Our results show that FLPA is generally preferable to LPA.

HiGHS is open-source optimization software for linear programming, mixed-integer programming, and quadratic programming. Created initially from research solvers written by Edinburgh PhD students, HiGHS attracted industrial funding that allowed further development, and saw it contribute to a REF 2021 Impact Case Study. Having been identified as a game-changer by the open-source energy systems planning community, the resulting crowdfunding campaign has received large donations that will allow the HiGHS project to expand and create further Impact.

This talk will give an insight into the state-of-the-art techniques underlying the linear programming solvers in HiGHS, with a particular focus on the challenge of solving sequences of linear systems of equations with remarkable properties. The means by which "gradware" created by PhD students has been transformed into software, generating income and Impact, will also be described. Independent benchmark results will be given to demonstrate that HiGHS is the world’s best open-source linear optimization software.

 

Residual networks (ResNets) have displayed impressive results in pattern recognition and, recently, have garnered considerable theoretical interest due to a perceived link with neural ordinary differential equations (neural ODEs). This link relies on the convergence of network weights to a smooth function as the number of layers increases. We investigate the properties of weights trained by stochastic gradient descent and their scaling with network depth through detailed numerical experiments. We observe the existence of scaling regimes markedly different from those assumed in neural ODE literature. Depending on certain features of the network architecture, such as the smoothness of the activation function, one may obtain an alternative ODE limit, a stochastic differential equation (SDE) or neither of these. Furthermore, we are able to formally prove the linear convergence of gradient descent to a global optimum for the training of deep residual networks with constant layer width and smooth activation function. We further prove that if the trained weights, as a function of the layer index, admit a scaling limit as the depth increases, then the limit has finite 2-variation.

In this talk, I will first give a brief introduction to the Landau-Ginzburg -- Conformal Field Theory (LG-CFT) correspondence, a prediction from physics. This prediction links aspects of Landau-Ginzburg models, described by matrix factorisations for a polynomial known as the potential, with Conformal Field Theories, described by for example vertex operator algebras. While both sides of the correspondence have good mathematical descriptions, it is an open problem to give a mathematical formulation of the correspondence. 

After this introduction, I will discuss the only known realisation of this correspondence, for the potential $x^d$. For even $d$ this is a recent result, and I will give a sketch of the proof which uses the tools of module tensor categories

 I will not assume prior knowledge of matrix factorisations, CFTs, or module tensor categories. This talk is based on joint work with Ana Ros Camacho.

In this talk, I will present several results on the Anderson Hamiltonian with white noise potential in dimension 1. This operator formally writes « - Laplacian + white noise ». It arises as the scaling limit of various discrete models and its explicit potential allows for a detailed description of its spectrum. We will discuss localization of its eigenfunctions as well as the behavior of the local statistics of its eigenvalues. Around large energies, we will see that the eigenfunctions are localized and follow a universal shape given by the exponential of a Brownian motion plus a drift, a behavior already observed by Rifkind and Virag in tridiagonal matrix models. Based on joint works with Cyril Labbé.

Shrinking Ricci solitons are Ricci flow solutions that self-similarly shrink under the flow. Their significance comes from the fact that finite-time Ricci flow singularities are typically modeled on gradient shrinking Ricci solitons. Here, we shall address a certain converse question, namely, “Given a complete, noncompact gradient shrinking Ricci soliton, does there exist a Ricci flow on a closed manifold that forms a finite-time singularity modeled on the given soliton?” We’ll discuss work that shows the answer is yes when the soliton is asymptotically conical. No symmetry or Kahler assumption is required, and so the proof involves an analysis of the Ricci flow as a nonlinear degenerate parabolic PDE system in its full complexity. We’ll also discuss applications to the (non-)uniqueness of weak Ricci flows through singularities.

Numerical optimization is an indispensable tool of modern data analysis, and there are many optimization problems where it is difficult or impossible to compute the full gradient of the objective function. The field of derivative free optimization (DFO) addresses these cases by using only function evaluations, and has wide-ranging applications from hyper-parameter tuning in machine learning to PDE-constrained optimization.

We present two projects that attempt to scale DFO techniques to higher dimensions.  The first method converges slowly but works in very high dimensions, while the second method converges quickly but doesn't scale quite as well with dimension.  The first-method is a family of algorithms called "stochastic subspace descent" that uses a few directional derivatives at every step (i.e. projections of the gradient onto a random subspace). In special cases it is related to Spall's SPSA, Gaussian smoothing of Nesterov, and block-coordinate descent. We provide convergence analysis and discuss Johnson-Lindenstrauss style concentration.  The second method uses conventional interpolation-based trust region methods which require large ill-conditioned linear algebra operations.  We use randomized linear algebra techniques to ameliorate the issues and scale to larger dimensions; we also use a matrix-free approach that reduces memory issues.  These projects are in collaboration with David Kozak, Luis Tenorio, Alireza Doostan, Kevin Doherty and Katya Scheinberg.

I will describe a procedure, known as holomorphic twist, to isolate protected quantities in supersymmetric quantum field theories. The resulting theories are holomorphic, interacting and have infinite dimensional symmetries, analogous to the holomorphic half of a 2D CFT. I will explain how to study quantum corrections to these symmetries and other  higher operations.
As a surprise, we find a novel UV manifestation of
confinement, dubbed "holomorphic confinement," in the example of pure
SU(N) super Yang-Mills.

Ilia Smilga
Margulis spacetimes and crooked planes

We are interested in the following problem: which groups can act 
properly on R^n by affine transformations, or in other terms, can occur 
as a symmetry group of a "regular affine tiling"? If we additionally 
require that they preserve a Euclidean metric (i.e. act by affine 
isometries), then these groups are well-known: they all contain a 
finite-index abelian subgroup. If we remove this requirement, a 
surprising result due to Margulis is that the free group can act 
properly on R^3. I shall explain how to construct such an action.

Charles Parker
Unexpected Behavior in Finite Elements for Linear Elasticity
One of the first problems that finite elements were designed to approximate is the small deformations of a linear elastic body; i.e. the 2D/3D version of Hooke's law for springs from elementary physics. However, for nearly incompressible materials, such as rubber, certain finite elements seemingly lose their approximation power. After briefly reviewing the equations of linear elasticity and the basics of finite element methods, we will spend most of the time looking at a few examples that highlight this unexpected behavior. We conclude with a theoretical result that (mostly) explains these findings.

The recent curation of large-scale databases with 3D surface scans of shapes has motivated the development of tools that better detect global patterns in morphological variation. Studies which focus on identifying differences between shapes have been limited to simple pairwise comparisons and rely on pre-specified landmarks (that are often known). In this talk, we present SINATRA: a machine learning pipeline for analyzing collections of shapes without requiring any correspondences. Our method takes in two classes of shapes and highlights the physical features that best describe the variation between them. We develop a rigorous simulation framework to assess our approach, which themselves are a novel contribution to 3D image and shape analyses. Lastly, as case studies with real data, we use SINATRA to (1) analyze mandibular molars from four different suborders of primates and (2) facilitate the visual identification of structural signatures differentiating between the trajectories of two protein ensembles resulting from molecular dynamics simulations. Together, these results highlight a promising future for interpretable machine learning to facilitate the non-trivial task of pattern recognition in evolutionary and structural biology with substantially increased resolution.

2:30 -3.00 Will Turner (CDT Student, Imperial College London)

Topologies on unparameterised path space

The signature of a path is a non-commutative exponential introduced by K.T. Chen in the 1950s, and appears as a central object in the theory of rough paths developed by T. Lyons in the 1990s. For continuous paths of bounded variation, the signature may be realised as a sequence of iterated integrals, which provides a succinct summary for multimodal, irregularly sampled, time-ordered data. The terms in the signature act as an analogue to monomials for finite dimensional data: linear functionals on the signature uniformly approximate any compactly supported continuous function on unparameterised path space (Levin, Lyons, Ni 2013). Selection of a suitable topology on the space of unparameterised paths is then key to the practical use of this approximation theory. We present new results on the properties of several candidate topologies for this space. If time permits, we will relate these results to two classical models: the fixed-time solution of a controlled differential equation, and the expected signature model of Levin, Lyons, and Ni. This is joint work with Thomas Cass.

3.05 -3.35 Ross Zhang (CDT Student, University of Oxford)

Random vortex dynamics via functional stochastic differential equations

The talk focuses on the representation of the three-dimensional (3D) Navier-Stokes equations by a random vortex system. This new system could give us new numerical schemes to efficiently approximate the 3D incompressible fluid flows by Monte Carlo simulations. Compared with the 2D Navier-Stokes equation, the difficulty of the 3D Navier-Stokes equation lies in the stretching of vorticity. To handle the stretching term, a system of stochastic differential equations is coupled with a functional ordinary differential equation in the 3D random vortex system. Two main tools are developed to derive the new system: the first is the investigation of pinned diffusion measure, which describes the conditional distribution of a time reversal diffusion, and the second is a forward-type Feynman Kac formula for nonlinear PDEs, which utilizes the pinned diffusion measure to delicately overcome the time reversal issue in PDE. Although the main focus of the research is the Navier-stokes equation, the tools developed in this research are quite general. They could be applied to other nonlinear PDEs as well, thereby providing respective numerical schemes.

3.40 - 4.25pm Dr Cris Salvi (Imperial College London)

Signature kernel methods

Kernel methods provide a rich and elegant framework for a variety of learning tasks including supervised learning, hypothesis testing, Bayesian inference, generative modelling and scientific computing. Sequentially ordered information often arrives in the form of complex streams taking values in non-trivial ambient spaces (e.g. a video is a sequence of images). In these situations, the design of appropriate kernels is a notably challenging task. In this talk, I will outline how rough path theory, a modern mathematical framework for describing complex evolving systems, allows to construct a family of characteristic kernels on pathspace known as signature kernels. I will then present how signature kernels can be used to develop a variety of algorithms such as two-sample hypothesis and (conditional) independence tests for stochastic processes, generative models for time series and numerical methods for path-dependent PDEs.

4.30 Refreshments

Building computationally capable biological systems has long been an aim of synthetic biology. The potential utility of bio-computing devices ranges from biosafety and environmental applications to diagnosis and personalised medicine. Here we present work on the design of bacterial computers which use spatial patterning to process information. A computer is composed of a number of bacterial colonies which, inspired by patterning in embryo development, communicate using diffusible morphogen-like signals. A computation is programmed into the overall physical arrangement of the system by arranging colonies such that the resulting diffusion field encodes the desired function, and the output is represented in the spatial pattern displayed by the colonies. We first mathematically demonstrate the simple digital logic capability of single bacterial colonies and show how additional structure is required to build complex functions. Secondly, inspired by electronic design automation, an algorithm for designing optimal spatial circuits computing two-level digital logic functions is presented, extending the capability of our system to complex digital functions without significantly increasing the biological complexity. We implement experimentally a proof-of-principle system using engineered Escherichia coli interpreting diffusion fields formed from droplets of an inducer molecule. Our approach will open up new ways to perform biological computation, with applications in synthetic biology, bioengineering and biosensing. Ultimately, these computational bacterial communities will help us explore information processing in natural biological systems.

The canonical dimension is a fundamental integer-valued invariant that is attached to mod p representations of p-adic Lie groups. I will explain why it is both an asymptotic measure of growth, and an algebraic quantity strongly related to Krull dimension. We will survey algebraic tools that can be applied in its calculation, and describe results spanning the last twenty years. I'll present a new theorem and suggest its possible significance for the mod p local Langlands programme. 

Movie Me would like offer dynamical pricing for movie tickets, considering consumer’s demand for the movie, showtime and lead time before the show begins, such that the overall quantity of tickets sold is maximized. We encourage all interested party to join us and especially those interested in data science, optimization and mathematical finance.

If $Q(x_0,x_1,x_2)$ is a quadratic form, how many solutions, of size at most $B$, does $Q=0$ have? How does this depend on $Q$? We apply the answers to the surface $y_0 Q_0 +y_1 Q_1 = 0$ in $P^1 x P^2$. (Joint work with Dan Loughran.)