Past

The first studies of the friction at the base of glacier were done by the pioneers Weertman and Lliboutry in the 1950s, who proposed theories under assumptions that have not been questioned for decades. Among these assumptions, the most questionable are the 2d geometry of the bumps, the pure sliding at the ice-bed interface and the hypothesis of stationary water pressure. In this seminar, I will present recent works using local modelling of basal friction with the finite element method that explore the validity of the proposed friction laws when these assumptions do not hold any more. 

The Temperley-Lieb algebra is a diagrammatic algebra - defined on a basis of "string diagrams" with multiplication given by "joining the diagrams together".  It first arose as an algebra of operators in statistical mechanics but quickly found application in knot theory (where Jones used it to define his famed polynomial) and the representation theory of $sl_2$.  From the outset, the representation theory of the Temperley-Lieb algebra itself has been of interest from a physics viewpoint and in characteristic zero it is well understood.  In this talk we will explore the representation theory over mixed characteristic (i.e. over positive characteristic fields and specialised at a root of unity).  This gentle introduction will take the listener through the beautifully fractal-like structure of the algebras and their cell modules with plenty of examples.

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 or neither of these. These findings cast doubts on the validity of the neural ODE model as an adequate asymptotic description of deep ResNets and point to an alternative class of differential equations as a better description of the deep network limit.
 

We revisit rough paths and signatures from a geometric and "smooth model" perspective. This provides a lean framework to understand and formulate key concepts of the theory, including recent insights on higher-order translation, also known as renormalization of rough paths. This first part is joint work with C Bellingeri (TU Berlin), and S Paycha (U Potsdam). In a second part, we take a semimartingale perspective and more specifically analyze the structure of expected signatures when written in exponential form. Following Bonnier-Oberhauser (2020), we call the resulting objects signature cumulants. These can be described - and recursively computed - in a way that can be seen as unification of previously unrelated pieces of mathematics, including Magnus (1954), Lyons-Ni (2015), Gatheral and coworkers (2017 onwards) and Lacoin-Rhodes-Vargas (2019). This is joint work with P Hager and N Tapia.

We develop a trust-region method for minimizing the sum of a smooth term f and a nonsmooth term h, both of which can be nonconvex. Each iteration of our method minimizes a possibly nonconvex model of f+h in a trust region. The model coincides with f+h in value and subdifferential at the center. We establish global convergence to a first-order stationary point when f satisfies a smoothness condition that holds, in particular, when it has Lipschitz-continuous gradient, and h is proper and lower semi-continuous. The model of h is required to be proper, lower-semi-continuous and prox-bounded. Under these weak assumptions, we establish a worst-case O(1/ε^2) iteration complexity bound that matches the best known complexity bound of standard trust-region methods for smooth optimization. We detail a special instance in which we use a limited-memory quasi-Newton model of f and compute a step with the proximal gradient method, resulting in a practical proximal quasi-Newton method. We describe our Julia implementations and report numerical results on inverse problems from sparse optimization and signal processing. Our trust-region algorithm exhibits promising performance and compares favorably with linesearch proximal quasi-Newton methods based on convex models.

This is joint work with Aleksandr Aravkin and Robert Baraldi.

-

A link for this talk will be sent to our mailing list a day or two in advance.  If you are not on the list and wish to be sent a link, please contact @email.

Noise is generated in an aerodynamic setting when flow turbulence encounters a structural edge, such as at the sharp trailing edge of an aerofoil. The generation of this noise is unavoidable, however this talk addresses various ways in which it may be mitigated through altering the design of the edge. The alterations are inspired by natural silent fliers: owls. A short review of how trailing-edge noise is modelled will be given, followed by a discussion of two independent adaptations; serrations, and porosity. The mathematical impacts of the adaptations to the basic trailing-edge model will be presented, along with the physical implications they have on noise generation and control.

In this talk I will introduce the study of lattices in locally compact groups through their actions CAT(0) spaces. This is an extremely rich class of groups including S-arithmetic groups acting on products of symmetric spaces and buildings, right angled Artin and Coxeter groups acting on polyhedral complexes, Burger-Mozes simple groups acting on products of trees, and the recent CAT(0) but non biautomatic groups of Leary and Minasyan. If time permits I will discuss some of my recent work related to the Leary-Minasyan groups.

There are many open conjectures about the algebraic behaviour of transcendental functions in arithmetic geometry, one of which is the Existential Closedness problem. In this talk I will review recent developments made on this question: the cases where we have unconditional existence of solutions, the conditional existence of generic solutions (depending on the conjecture of periods and Zilber-Pink), and even a few cases of unconditional existence of generic solutions. Many of the results I will mention are joint work with (different subsets of) Vahagn Aslanyan, Jonathan Kibry, Sebastián Herrero, and Roy Zhao. 

https://teams.microsoft.com/l/meetup-join/19%3ameeting_ZGRiMTM1ZjQtZWNi…

I will introduce the notion of moduli spaces of curves and specifically genus 0 curves. They are in general not compact, and we will discuss the most common way to compactify them. In particular, I will try to explain the construction of Mbar_{0,5}, together with how to classify the boundary, how it is related to a moduli space of tropical curves, and how to do intersection theory on this space.

In this talk, I will discuss AdS super gluon scattering amplitudes in various spacetime dimensions. These amplitudes are dual to correlation functions in a variety of non-maximally supersymmetric CFTs, such as the 6d E-string theory, 5d Seiberg exceptional theories, etc. I will introduce a powerful method based on symmetries and consistency conditions, and show that it fixes all the infinitely many four-point amplitudes at tree level. I will also point out many interesting properties and structures of these amplitudes, which include the flat space limit, Parisi-Sourlas-like dimensional reduction, hidden conformal symmetry, and a color-kinematic duality in AdS. Along the way, I will also review some earlier progress and the relation with this work. I will conclude with a brief discussion of various open problems. 

A group of projection based approaches for solving large-scale linear systems is known for its speed and simplicity. For example, Kaczmarz algorithm iteratively projects the previous approximation x_k onto the solution spaces of the next equation in the system. An elegant proof of the exponential convergence of this method, using correct randomization of the process, was given in 2009 by Strohmer and Vershynin, and succeeded by many extensions and generalizations. I will discuss our newly developed variants of these methods that successfully avoid large and potentially adversarial corruptions in the linear system. I specifically focus on the random matrix and high-dimensional probability results that play a crucial role in proving convergence of such methods. Based on the joint work with Jamie Haddock, Deanna Needell, and Will Swartworth.

We discuss the behavior of geodesics in the continuous models of random geometry known as the Brownian map and the Brownian plane. We say that a point $x$ is a geodesic star with $m$ arms if $x$ is the endpoint of $m$ disjoint geodesics. We prove that the set of all geodesic stars with $m$ arms has dimension $5-m$, for $m=1,2,3,4$. This complements recents results of Miller and Qian, who derived upper bounds for these dimensions.

Riemannian optimization is a powerful and active area of research that studies the optimization of functions defined on manifolds with structure. A class of functions of interest is the set of geodesically convex functions, which are functions that are convex when restricted to every geodesic. In this talk, we will present an accelerated first-order method, nearly achieving the same rates as accelerated gradient descent in the Euclidean space, for the optimization of smooth and g-convex or strongly g-convex functions defined on the hyperbolic space or a subset of the sphere. We will talk about accelerated optimization of another non-convex problem, defined in the Euclidean space, that we solve as a proxy. Additionally, for any Riemannian manifold of bounded sectional curvature, we will present reductions from optimization methods for smooth and g-convex functions to methods for smooth and strongly g-convex functions and vice versa.

This talk is based on the paper https://arxiv.org/abs/2012.03618.

-

A link for this talk will be sent to our mailing list a day or two in advance.  If you are not on the list and wish to be sent a link, please contact @email.

How much does the chromatic number of the random graph $G(n, 1/2)$ vary? Shamir and Spencer proved that it is contained in some sequence of intervals of length about $n^{1/2}$. Alon improved this slightly to $n^{1/2} / \log n$. Until recently, however, no lower bounds on the fluctuations of the chromatic number of $G(n, 1/2)$ were known, even though the question was raised by Bollobás many years ago. I will talk about the main ideas needed to prove that, at least for infinitely many $n$, the chromatic number of $G(n, 1/2)$ is not concentrated on fewer than $n^{1/2-o(1)}$ consecutive values.
I will also discuss the Zigzag Conjecture, made recently by Bollobás, Heckel, Morris, Panagiotou, Riordan and Smith: this proposes that the correct concentration interval length 'zigzags' between $n^{1/4+o(1)}$ and $n^{1/2+o(1)}$, depending on $n$.
Joint work with Oliver Riordan.

The complementarity and substitutability between products are essential concepts in retail and marketing. Qualitatively, two products are said to be substitutable if a customer can replace one product by the other, while they are complementary if they tend to be bought together. In this article, we take a network perspective to help automatically identify complements and substitutes from sales transaction data. Starting from a bipartite product-purchase network representation, with both transaction nodes and product nodes, we develop appropriate null models to infer significant relations, either complements or substitutes, between products, and design measures based on random walks to quantify their importance. The resulting unipartite networks between products are then analysed with community detection methods, in order to find groups of similar products for the different types of relationships. The results are validated by combining observations from a real-world basket dataset with the existing product hierarchy, as well as a large-scale flavour compound and recipe dataset.

arXiv link: https://arxiv.org/abs/2103.02042

We propose a randomized algorithm for solving a linear system $Ax = b$ with a highly numerically rank-deficient coefficient matrix $A$ with nearly consistent right-hand side possessing a small-norm solution. Our algorithm finds a small-norm solution with small residual in $O(N_r + nrlogn + r^3 )$ operations, where $r$ is the numerical rank of $A$ and $N_r$ is the cost of multiplying an $n\times r$ matrix to $A$. 

Joint work with Marcus Webb (Manchester). 

A link for this talk will be sent to our mailing list a day or two in advance.  If you are not on the list and wish to be sent a link, please contact @email.

The index of a saddle point of a smooth function is the number of descending directions of the saddle. While the index can usually be retrieved by counting the number of negative eigenvalues of the Hessian at the critical point, we may not have the luxury of having second derivatives in data deriving from practical applications. To address this problem, we develop a computational pipeline for estimating the index of a non-degenerate saddle point without explicitly computing the Hessian. In our framework, we only require a sufficiently dense sample of level sets of the function near the saddle point. Using techniques in Morse theory and Topological Data Analysis, we show how the shape of saddle points can help us infer the index of the saddle. Furthermore, we derive an explicit upper bound on the density of point samples necessary for inferring the index depending on the curvature of level sets. 

The outstanding issue of a non-singular extension of the Kerr-NUT- (anti) de Sitter solutions to Einstein’s equations is solved completely. The Misner’s method of obtaining the extension for Taub-NUT spacetime is generalized in a non-singular manner. The Killing vectors that define non-singular spaces of non-null orbits are derived and applied. The global structure of spacetime is discussed. The non-singular conformal geometry of theinfinities is derived. The Killing horizons are present.

A fundamental question in fluid dynamics concerns the formation of discontinuous shock waves from smooth initial data. In previous works, we have established stable generic shock formation for the compressible Euler system, showing that at the first singularity the solution has precisely C^{1/3} Holder regularity, a so-called preshock. The focus of this talk is a complete space-time description of the solution after this initial singularity. We show that three surfaces of discontinuity emerge simultaneously and instantaneously from the preshock: the classical shock discontinuity that propagates by the Rankine–Hugoniot conditions, together with two distinct surfaces in space-time, along which C^{3/2} cusp singularities form.

For every positive integer N and every α ∈ [0,1), let B(N, α) denote the probabilistic model in which a random set A of (1,...,N) is constructed by choosing independently every element of (1,...,N) with probability α. We prove that, as N → +∞, for every A in B(N, α) we have |AA| ~ |A|^2/2 with probability 1-o(1), if and only if (log(α^2(log N)^{log 4-1}))(√loglog N) → ∞. This improves on a theorem of Cilleruelo, Ramana and Ramar\'e, who proved the above asymptotic between |AA| and |A|^2/2 when α =o(1/√log N), and supplies a complete characterization of maximal product sets of random sets.

The unknotting number of a knot is the minimum number of crossing changes needed to untie the knot. It is one of the simplest knot invariants to define, yet remains notoriously difficult to compute. We will survey some basic knot theory invariants and constructions, including the satellite knot construction, a straightforward way to build new families of knots. We will give a lower bound on the unknotting number of certain satellites using knot Floer homology. This is joint work in progress with Tye Lidman and JungHwan Park.

The Atiyah-Floer conjecture asserts the instanton Floer homology of a closed three-manifold (constructed via gauge theory) is isomorphic to the Lagrangian Floer homology of a pair of Lagrangian submanifolds associated to a splitting of the three manifold (constructed via symplectic geometry). This conjecture has remained open for more than three decades. In this talk I will explain two compactness results for the SO(3) case of the conjecture in the neck-stretching process. One result is related to the construction of a natural bounding chain in the Lagrangian Floer theory and a conjecture of Fukaya.

We discuss constraints from perturbative unitarity and crossing on the leading contributions of the higher-dimension operators to the four-graviton amplitude in four spacetime dimensions. We focus on the leading order effect due to exchange by massive degrees of freedom which makes the amplitudes of interest IR finite. To test the constraints we obtain nontrivial effective field theory data by computing and taking the large mass expansion of the one-loop minimally-coupled four-graviton amplitude with massive particles up to spin 2 circulating in the loop. Remarkably, the leading EFT corrections to Einstein gravity of physical theories, both string theory and QFT coupled to gravity, end up in minuscule islands which are much smaller than what is suggested by the generic bounds obtained from consistency of the 2-2 graviton scattering amplitude. We discuss the underlying mechanism for this phenomenon.