L4

Domain decomposition methods are widely employed for the numerical solution of partial differential equations on parallel computers. We develop an adjoint-based a posteriori error analysis for overlapping multiplicative Schwarz domain decomposition and for overlapping additive Schwarz. In both cases the numerical error in a user-specified functional of the solution (quantity of interest), is decomposed into a component that arises due to the spatial discretization and a component that results from of the finite iteration between the subdomains. The spatial discretization error can be further decomposed in to the errors arising on each subdomain. This decomposition of the total error can then be used as part of a two-stage approach to construct a solution strategy that efficiently reduces the error in the quantity of interest.

Can we get rigorous answers when computing with real and complex numbers? There are now many applications where this is possible thanks to a combination of tools from computer algebra and traditional numerical computing. I will give an overview of such methods in the context of two projects I'm developing. The first project, Arb, is a library for arbitrary-precision ball arithmetic, a form of interval arithmetic enabling numerical computations with rigorous error bounds. The second project, Fungrim, is a database of knowledge about mathematical functions represented in symbolic form. It is intended to function both as a traditional reference work and as a software library to support symbolic-numeric methods for problems involving transcendental functions. I will explain a few central algorithmic ideas and explain the research goals of these projects.

The classical Universal Approximation Theorem certifies that the universal approximation property holds for the class of neural networks of arbitrary width. Here we consider the natural `dual' theorem for width-bounded networks of arbitrary depth, for a broad class of activation functions. In particular we show that such a result holds for polynomial activation functions, making this genuinely different to the classical case. We will then discuss some natural extensions of this result, e.g. for nowhere differentiable activation functions, or for noncompact domains.
 

The hard Lefschetz theorem is a fundamental statement about the symmetry of the cohomology of algebraic varieties. In nearly all cases that we systematically understand it, it comes with a geometric meaning, often in form of Hodge structures and signature data for the Hodge-Riemann bilinear form.

Nevertheless, similar to the role the standard conjectures play in number theory, several intriguing combinatorial problems can be reduced to hard Lefschetz properties, though in extreme cases without much geometric meaning, lacking any existence of, for instance,  an ample cone to do Hodge theory with.

I will present a way to prove the hard Lefschetz theorem in such a situation, by introducing biased pairing and perturbation theory for intersection rings. The price we pay is that the underlying variety, in a precise sense, has itself to be sufficiently generic. For instance, we shall see that any quasismooth, but perhaps nonprojective toric variety can be "perturbed" to a toric variety with the same equivariant cohomology, and that has the hard Lefschetz property.

Finally, I will discuss how this applies to prove some interesting theorems in geometry, topology and combinatorics. In particular, we shall see a generalization of a classical result due to Descartes and Euler: We prove that if a simplicial complex embeds into euclidean 2d-space, the number of d-simplices in it can exceed the number of (d-1)-simplices by a factor of at most d+2.

A fundamental problem in the context of Einstein's equations of general relativity is to understand precisely the dynamical evolution of small perturbations of stationary black hole solutions. It is expected that there is a discrete set of characteristic frequencies that play a dominant role at late time intervals and carry information about the nature of the black hole, much like the normal frequencies of a vibrating string. These frequencies are called quasi-normal frequencies or resonances and they are closely related to scattering resonances in the study of Schrödinger-type equations. I will discuss a new method of defining and studying resonances for linear wave equations on asymptotically flat black holes, developed from joint work with Claude Warnick.

Machine learning for scientific applications faces the challenge of limited data. To reduce overfitting, we propose a framework to leverage as much as possible a priori-known physics for a problem. Our approach embeds a deep neural network in a partial differential equation (PDE) system, where the pre-specified terms in the PDE capture the known physics and the neural network will learn to describe the unknown physics. The neural network is estimated from experimental and/or high-fidelity numerical datasets. We call this approach a “deep learning PDE model” (DPM). Once trained, the DPM can be used to make out-of-sample predictions for new physical coefficients, geometries, and boundary conditions. We implement our approach on a classic problem of the Navier-Stokes equation for turbulent flows. The numerical solution of the Navier-Stokes equation (with turbulence) is computationally expensive and requires a supercomputer. We show that our approach can estimate a (computationally fast) DPM for the filtered velocity of the Navier-Stokes equations. 

The Anderson Hamiltonian is used to model particles moving in
disordered media, it can be thought of as a Schrödiger operator with an
extremely irregular random potential. Using the recently developed theory of
"Paracontrolled Distributions" we are able to define the Anderson
Hamiltonian as a self-adjoint non-positive operator on the 2- and
3-dimensional torus and give an explicit description of its domain.
Then we use these results to solve some semi-linear PDEs whose linear part
is given by the Anderson Hamiltonian, more precisely the multiplicative
stochastic NLS and nonlinear Wave equation.
This is joint work with M. Gubinelli and B. Ugurcan.

In this talk I will explain (a) what observations speak for the
hypothesis of dark matter, (b) what observations speak for
the hypothesis of modified gravity, and (c) why it is a mistake
to insist that either hypothesis on its own must
explain all the available data. The right explanation, I will argue,
is instead a suitable combination of dark matter and modified
gravity, which can be realized by the idea that dark matter
has a superfluid phase.

Magnetic skyrmions are topological solitons which occur in a large class
of ferromagnetic materials and which are currently attracting much
attention in the condensed matter community because of  their possible
use  in future magnetic information storage technology.  The talk is
about an integrable model for magnetic skyrmions, introduced in a recent
paper (arxiv 1812.07268) and generalised in (arxiv 1905.06285). The
model can be solved by interpreting it as a gauged nonlinear sigma
model. In the talk will explain the model and the geometry behind its
integrability, and discuss some of the solutions and their physical
interpretation.