Past

In this talk, we present a topological framework for interpreting the latent representations of Multilayer Perceptrons (MLPs) [1] using tools from Topological Data Analysis. Our approach constructs a simplicial tower, a sequence of simplicial complexes linked by simplicial maps, to capture how the topology of data evolves across network layers. This construction is based on the pullback of a cover tower on the output layer and is inspired by the Multiscale Mapper algorithm. The resulting commutative diagram enables a dual analysis: layer persistence, which tracks topological features within individual layers, and MLP persistence, which monitors how these features transform across layers. Through experiments on both synthetic and real-world medical datasets, we demonstrate how this method reveals critical topological transitions, identifies redundant layers, and provides interpretable insights into the internal organization of neural networks.

Let p be a prime. In this talk we look at the bounded derived category of modules over the Rubik’s cube group and show that the faithful action on the corners and edges is a progenerator for the coadmissible subcategory.

The motion of a particle suspended in a fluid flow is governed by hydrodynamic interactions. In this talk, I will present the rich nonlinear dynamics that arise from particle-fluid interactions for two different setups: (i) passive particles in 3D channel flows where fluid inertia is important, and (ii) active particles in 3D channel flows in the Stokes regime (i.e. without fluid inertia).

For setup (i), the particle-fluid interactions result in focusing of particles in the channel cross section, which has been exploited in biomedical microfluidic technologies to separate particles by size. I will offer insights on how dynamical system features of bifurcations and tipping phenomena might be exploited to efficiently separate particles of different sizes. For setup (ii), microswimmers routinely experience unidirectional flows in confined environment such as sperm cells swimming in fallopian tubes, pathogens moving through blood vessels, and microrobots programed for targeted drug delivery applications. I will show that our minimal model of the system exhibits rich nonlinear and chaotic dynamics resulting in a diverse set of active particle trajectories.

Odd systems, characterised by broken time-reversal or parity symmetry, 
exhibit striking transport phenomena due to transverse responses. In this 
talk, I will introduce the concept of odd diffusion, a generalisation of 
diffusion in two-dimensional systems that incorporates antisymmetric tensor 
components. Focusing on systems of interacting particles, I adapt a 
geometric approach to derive effective transport equations and show how 
interactions give rise to unusual transport in odd systems. I present 
effects like enhanced self-diffusion, reversed Hall drift and even absolute 
negative mobility that solely originate in odd diffusion. These results 
reveal how microscopic symmetry-breaking gives rise to emergent, equilibrium 
and non-equilibrium transport, with implications for soft matter, chiral 
active systems, and topological materials.

We introduce a non-probabilistic, path-by-path framework for continuous-time, path-dependent portfolio allocation. Extending the self-financing concept recently introduced in Chiu & Cont (2023), we characterize self-financing portfolio allocation strategies through a path-dependent PDE and provide explicit solutions for the portfolio value in generic markets, including price paths that are not necessarily continuous or exhibit variation of any order.

As an application, we extend an aggregating algorithm of Vovk and the universal algorithm of Cover to continuous-time meta-algorithms that combine multiple strategies into a single strategy, respectively tracking the best individual and the best convex combination of strategies. This work extends Cover’s theorem to continuous-time without probability.

Let $p$ be an odd prime. Let $K/\mathbf{Q}_p$ be a finite unramified extension. Let $\rho: G_K \to \mathrm{GL}_2(\overline{\mathbf{F}}_p)$ be a continuous representation. We prove that $\rho$ has a crystalline lift of small irregular weight if and only if it has multiple crystalline lifts of certain specified regular weights. The inspiration for this result comes from recent work of Diamond and Sasaki on geometric Serre weight conjectures. We also discuss applications to partial weight one modularity.

Ever wondered how ideas from physics can used in real-world scenarios? Come to this talk to understand what is an option and how they are traded in markets. I will recall some basic notions of stochastic calculus and derive the Black-Scholes (BS) equation for plain vanilla options. The BS equation can be solved using standard path integral techniques, that also allow to price more exotic derivatives. Finally, I will discuss whether the assumptions behind Black-Scholes dynamics are reasonable in real-world markets (spoiler: they're not), volatility smiles and term structures of the implied volatility.

Junior Strings is a seminar series where DPhil students present topics of common interest that do not necessarily overlap with their own research area. This is primarily aimed at PhD students and post-docs but everyone is welcome.

We will discuss local regularity properties for solutions to non-local equations naturally arising in kinetic theory. We will focus on the Strong Harnack inequality for global solutions to a non-local kinetic equation in divergence form. We will explain the connection to the Boltzmann equation and we will mention a few consequences on the asymptotic behaviour of the solutions.

Building on Leo’s talk last week, I will present the full Galois characterisation of henselianity and introduce some of the ‘explicit’ ingredients he referred to during his presentation. In particular, I will describe a Galois cohomology-inspired criterion for distinguishing between different characteristics. I will then outline the full proof of the Galois characterisation of p-adically closed fields, indicating how each of the ingredients enters the argument.

Is it possible to tell the isomorphism type of an infinite group from its collection of finite quotients? This question, known as profinite rigidity, has deep roots in various areas of mathematics, ranging from arithmetic geometry to group theory. In this talk, I will introduce the question, its history and context. I will explain how profinite rigidity is studied using the machinery of profinite completions, including elementary proofs and counterexamples. Then I will outline some of the key results in the field, ranging from 1970 to the present day. Time permitting, I will elaborate on recent results of myself on the profinite rigidity of certain classes of solvable groups. 

Structural mechanics plays a crucial role in soft matter physics, mechanobiology, metamaterials, pattern formation, active matter, and soft robotics. What unites these seemingly disparate topics is the natural balance that emerges between elasticity, geometry, and stability. This seminar will serve as a high-level overview of our work on several problems concerning the stability of structures. I will cover three topics: (1) shapeshifting shells; (2) mechanical metamaterials; and (3) elastogranular mechanics.

I will begin by discussing our development of a generalized, stimuli-responsive shell theory. (1) Non-mechanical stimuli including heat, swelling, and growth further complicate the nonlinear mechanics of shells, as simultaneously solving multiple field equations to capture multiphysics phenomena requires significant computational expense. We present a general shell theory to account for non-mechanical stimuli, in which the effects of the stimuli are
generalized into three forms: those that add mass to the shell, those that increase the area of the shell through the natural stretch, and those that change the curvature of the shell through the natural curvature. I will show how this model can capture the morphogenesis of the optic cup, the snapping of the Venus flytrap, leaf growth, and the buckling of electrically active polymer plates. (2) I will then discuss how cutting thin sheets and shells, a process
inspired by the art of kirigami, enables the design of functional mechanical metamaterials. We create linear actuators, artificial muscles, soft robotic grippers, and mechanical logic units by systematically cutting and stretching thin sheets. (3) Finally, if time permits, I will introduce our work on the interactions between elastic and granular matter, which we refer to as elastogranular mechanics. Such interactions occur across all lengths, from morphogenesis, to root growth, to stabilizing soil against erosion. We show how combining rocks and string in the absence of any adhesive we can create large, load bearing structures like columns, beams, and arches. I will finish with a general phase diagram for elastogranular behavior.

I will describe a measure of quantum state complexity defined by minimizing the spread of the wavefunction over all choices of basis. We can efficiently compute this measure, which displays universal behavior for diverse chaotic systems including spin chains, the SYK model, and quantum billiards.  In the minimizing basis, the Hamiltonian is tridiagonal, thus representing the dynamics as if they unfold on a one-dimensional chain. The recurrent and hopping matrix elements of this chain comprise the Lanczos coefficients, which I will relate through an integral formula to the density of states. For Random Matrix Theories (RMTs), which are believed to describe the energy level statistics of chaotic systems, I will also derive an integral formula for the covariances of the Lanczos coefficients. These results lead to a conjecture: quantum chaotic systems have Lanczos coefficients whose local means and covariances are described by RMTs. 
 

The hydrodynamic description for emergent behavior of interacting agents is governed by Euler alignment equations, driven by different protocols of pairwise communication kernels. A main question of interest is how short- vs. long-range interactions dictate the large-crowd, long-time dynamics. 

The equations lack closure for the pressure away thermal equilibrium. We identify a distinctive feature of Euler alignment -- a reversed direction of entropy. We discuss the role of a reversed entropy inequality in selecting mono-kinetic closure for emergence of strong solutions, prove the existence of such solutions, and characterize their related invariants which extend the 1-D notion of an “e” quantity.

The hydrodynamic description for emergent behavior of interacting agents is governed by Euler alignment equations, driven by different protocols of pairwise communication kernels. A main question of interest is how short- vs. long-range interactions dictate the large-crowd, long-time dynamics. 

The equations lack closure for the pressure away thermal equilibrium. We identify a distinctive feature of Euler alignment -- a reversed direction of entropy. We discuss the role of a reversed entropy inequality in selecting mono-kinetic closure for emergence of strong solutions, prove the existence of such solutions, and characterize their related invariants which extend the 1-D notion of an “e” quantity.

Roe algebras were introduced in the late 1990's in the study of indices of elliptic operators on (locally compact) Riemannian manifolds. Roe was particularly interested in coarse equivalences of metric spaces, which is a weaker notion than that of quasi-isometry. In fact, soon thereafter it was realized that the isomorphism class of these class of C*-algebras did not depend on the coarse equivalence class of the manifold. The converse, that is, whether this class is a complete invariant, became known as the 'Rigidity Problem for Roe algebras'. In this talk we will discuss an affirmative answer to this question, and how to approach its proof. This is based on joint work with Federico Vigolo.

Quiver Donaldson-Thomas invariants are integers determined by the geometry of moduli spaces of quiver representations. I will describe a correspondence between quiver Donaldson-Thomas invariants and Gromov-Witten counts of rational curves in toric and cluster varieties. This is joint work with Pierrick Bousseau.

Lagrangian mean curvature flow (LMCF) is a way to deform Lagrangian submanifolds inside a Calabi-Yau manifold according to the negative gradient of the area functional. There are influential conjectures about LMCF due to Thomas-Yau and Joyce, describing the long-time behaviour of the flow, singularity formation, and how one may flow past singularities. In this talk, we will show how to flow past a conically singular Lagrangian by gluing in expanders asymptotic to the cone, generalizing an earlier result by Begley-Moore. We solve the problem by a direct P.D.E.-based approach, along the lines of recent work by Lira-Mazzeo-Pluda-Saez on the network flow. The main technical ingredient we use is the notion of manifolds with corners and a-corners, as introduced by Joyce following earlier work of Melrose.

Do there exist universal sequences for all mazes on the two-dimensional integer lattice? We will give background on this question, as well as some recent results. Joint work with Mariaclara Ragosta.

Let X be a smooth rigid-analytic variety. Ardakov and Wadsley introduced the sheaf D-cap of infinite order differential operators on X, along with the category of coadmissible D-cap-modules. In this talk, we present a Riemann–Hilbert correspondence for these coadmissible D-cap-modules. Specifically, we interpret a coadmissible D-cap-module as a p-adic differential equation, explain what it means to solve such an equation, and describe how to reconstruct the module from its solutions.

Einstein’s equations are difficult to solve and if you want to compute something in holography knowing an explicit metric seems to be essential. Or is it? For some theories, observables, such as on-shell actions and free energies, are determined solely in terms of topological data, and an explicit metric is not needed. One of the key tools that has recently been used for this programme is equivariant localization, which gives a method of computing integrals on spaces with a symmetry. In this talk I will give a pedestrian introduction to equivariant localization before showing how it can be used to compute the on-shell action of 6d Romans Gauged supergravity.