Past

The Drinfeld half-plane is a rigid analytic variety over a p-adic field. In this talk, I will give an overview of the geometric aspects of this space and describe its connection with representation theory.

We develop a framework for showing that neural networks can overcome the curse of dimensionality in different high-dimensional approximation problems. Our approach is based on the notion of a catalog network, which is a generalization of a standard neural network in which the nonlinear activation functions can vary from layer to layer as long as they are chosen from a predefined catalog of functions. As such, catalog networks constitute a rich family of continuous functions. We show that under appropriate conditions on the catalog, catalog networks can efficiently be approximated with ReLU-type networks and provide precise estimates on the number of parameters needed for a given approximation accuracy. As special cases of the general results, we obtain different classes of functions that can be approximated with ReLU networks without the curse of dimensionality. 

A preprint is here: https://arxiv.org/abs/1912.04310

Let G be a reductive group, E an elliptic curve, and Bun_G the moduli stack of principal G-bundles on E. In this talk, I will attempt to explain why Bun_G is a very interesting object from the perspectives of both singularity theory on the one hand, and shifted symplectic geometry and representation theory on the other. In the first part of the talk, I will explain how to construct slices of Bun_G through points corresponding to unstable bundles, and how these are linked to certain singular algebraic surfaces and their deformations in the case of a "subregular" bundle. In the second (probably much shorter) part, I will discuss the shifted symplectic geometry of Bun_G and its slices. If time permits, I will sketch how (conjectural) quantisations of these structures should be related to some well known algebras of an "elliptic" flavour, such as Sklyanin and Feigin-Odesskii algebras, and elliptic quantum groups.

In this talk, we will review the notion of non-invertible symmetries and we will study adjoint QCD in two dimensions. It turns out that this theory has a plethora of such symmetries which require deconfinement in the massless case. When a mass or certain quartic interactions are tunrned on, these symmetries are broken and the theory confines. In addition, we will use these symmetries to calculate the string tension for small mass and make some comments about naturalness along the RG flow.

Research on robot manipulation has focused, in recent years, on grasping everyday objects, with target objects almost exclusively rigid items. Non–rigid objects, as textile ones, pose many additional challenges with respect to rigid object manipulation. In this seminar we will present how we can employ topology to study the ``state'' of a rectangular textile using the configuration space of $n$ points on the plane. Using a CW-decomposition of such space, we can define for any mesh associated with a rectangular textile a vector in an euclidean space with as many dimensions as the number of regions we have defined. This allows us to study the distribution of such points on the cloth and define meaningful states for detection and manipulation planning of textiles. We will explain how such regions can be defined and computationally how we can assign to any mesh the corresponding region. If time permits, we will also explain how the CW-structure allows us to define more than just euclidean distance between such mesh-distributions.

One of the key objects in studying the Hilbert Scheme of points in the plane is a torus action of $(\mathbb{C}^*)^2$. The fixed points of this action correspond to monomial ideals in $\mathbb{C}[x,y]$, and this gives a connection between the geometry of Hilbert schemes and partition combinatorics. Using this connection, one can extract identities in partition combinatorics from algebro-geometric information and vice versa. I will give some examples of combinatorial identities where as yet the only proofs we have rely on the geometry of Hilbert schemes. If there is time, I will also sketch out a hope that such identities can also be seen by representations of appropriately chosen algebras.

Convection is the main heat transport process in the liquid cores of planets and the primary energy source for planetary magnetic fields. These convective motions are thought to be turbulent and strongly constrained by rotation. In this talk, I will discuss the large-scale flows (zonal jets and vortices) that form in this rapidly-rotating turbulent regime, which we explore with numerical models.

Genotype-phenotype associations can be results of direct effects, genetic nurturing effects and population stratification confounding (The nature of nurture: Effects of parental genotypes, Science, 2018, Deconstructing the sources of genotype-phenotype associations in humans, Science, 2019). Genotypes from parents and siblings of the proband can be used to statistically disentangle these effects. To maximize power, a comprehensive framework for utilizing various combinations of parents’ and siblings’ genotypes is introduced. Central to the approach is mendelian imputation, a method that utilizes identity by descent (IBD) information to non-linearly impute genotypes into untyped relatives using genotypes of typed individuals. Applying the method to UK Biobank probands with at least one parent or sibling genotyped, for an educational attainment (EA) polygenic score that has a R² of 5.7% with EA, its predictive power based on direct genetic effect alone is demonstrated to be only about 1.4%. For women, the EA polygenic score has a bigger estimated direct effect on age-at-first-birth than EA itself.

TBA

Abstract: In many situations, one needs to decide between acting to reveal data about a system and acting to generate profit; this is the trade-off between exploration and exploitation. A simple situation where we face this trade-off is a multiarmed bandit problem, where one has M ‘bandits’ which generate reward from an unknown distribution, and one must choose which bandit to play at each time. The key difficulty in the multi-armed bandit problem is that the action often affects the information obtained. Due to the curse of dimensionality, solving the bandit problem directly is often computationally intractable.

In this talk, we will formulate a general class of the multi-armed bandit problem as a relaxed stochastic control problem. By introducing an entropy premium, we obtain a smooth asymptotic approximation to the value function. This yields a novel semi-index approximation of the optimal decision process, obtained numerically by solving a fixed point problem, which can be interpreted as explicitly balancing an exploration–exploitation trade-off.  Performance of the resulting Asymptotic Randomised Control (ARC) algorithm compares favourably with other approaches to correlated multi-armed bandits.

As an application of the multi-armed bandit, we also consider a multi-armed bandit problem where the observation from each bandit arrive from a Generalised Linear Model. We then use such model to consider a dynamic online pricing problem. The numerical simulation shows that the ARC algorithm also performs well compared to others.
============

There are many constructions that yield C*-algebras. For example, we build them from groups, quantum groups, dynamical systems, and graphs. In this talk we look at C*-algebras that arise from a certain type of game. It turns out that the properties of the underlying game gives us very strong information about existence of traces of various types on the game algebra. The recent solution of the Connes Embedding Problem arises from a game whose algebra has a trace but no hyperlinear trace.

Assumed knowledge: Familiarity with tensor products of Hilbert spaces, the algebra of a discrete group, and free products of groups.

Kirigami, the relatively unheralded cousin of origami, is the art of cutting paper to articulate and deploy it as a whole. By varying the number, size, orientation and coordination of the cuts, artists have used their imagination and intuition to create remarkable sculptures in 2 and 3 dimensions. I will describe some of our attempts to quantify the inverse problem that artists routinely solve, combining elementary mathematical ideas, with computations and physical models. 

[[{"fid":"60095","view_mode":"media_portrait_large","fields":{"format":"media_portrait_large","field_file_image_alt_text[und][0][value]":false,"field_file_image_title_text[und][0][value]":false},"type":"media","field_deltas":{"1":{"format":"media_portrait_large","field_file_image_alt_text[und][0][value]":false,"field_file_image_title_text[und][0][value]":false}},"attributes":{"class":"media-element file-media-portrait-large","data-delta":"1"}}]]

Reconstructing continuous signals based on a small number of discrete samples is a fundamental problem across science and engineering. In practice, we are often interested in signals with ``simple'' Fourier structure -- e.g., those involving frequencies within a bounded range, a small number of frequencies, or a few blocks of frequencies. More broadly, any prior knowledge about a signal's Fourier power spectrum can constrain its complexity.  Intuitively, signals with more highly constrained Fourier structure require fewer samples to reconstruct.

We formalize this intuition by showing that, roughly speaking, a continuous signal from a given class can be approximately reconstructed using a number of samples equal to the statistical dimension of the allowed power spectrum of that class. We prove that, in nearly all settings, this natural measure tightly characterizes the sample complexity of signal reconstruction.

Surprisingly, we also show that, up to logarithmic factors, a universal non-uniform sampling strategy can achieve this optimal complexity for any class of signals. We present a simple, efficient, and general algorithm for recovering a signal from the samples taken. For bandlimited and sparse signals, our method matches the state-of-the-art. At the same time, it gives the first computationally and sample efficient solution to a broad range of problems, including multiband signal reconstruction and common kriging and Gaussian process regression tasks.

Our work is based on a novel connection between randomized linear algebra and the problem of reconstructing signals with constrained Fourier structure. We extend tools based on statistical leverage score sampling and column-based matrix reconstruction to the approximation of continuous linear operators that arise in the signal fitting problem. We believe that these extensions are of independent interest and serve as a foundation for tackling a broad range of continuous time problems using randomized methods.

This is joint work with Michael Kapralov, Cameron Musco, Christopher Musco, Ameya Velingker and Amir Zandieh

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 send email to @email.

In 2010, Hrushovski--Loeser studied the homotopy type of the Berkovich analytification of a quasi-projective variety over a valued field. In this talk, we explore the extent to which some of their results might hold in a relative setting. More precisely, given a morphism of quasi-projective varieties over a valued field, we ask if we might construct deformation retractions of the analytifications of the source and target which are compatible with the analytification of the morphism and whose images are finite simplicial complexes. 

An inner model is a ground if V is a set forcing extension of it. The intersection of the grounds is the mantle, an inner model of ZFC which enjoys many nice properties. Fuchs, Hamkins, and Reitz showed that the mantle is highly malleable. Namely, they showed that every model of set theory is the mantle of a bigger, better universe of sets. This then raises the possibility of iterating the definition of the mantle—the mantle, the mantle of the mantle, and so on, taking intersections at limit stages—to obtain even deeper inner models. Let’s call the inner models in this sequence the inner mantles.

In this talk I will present some results, both positive and negative, about the sequence of inner mantles, answering some questions of Fuchs, Hamkins, and Reitz, results which are analogues of classic results about the sequence of iterated HODs. On the positive side: (Joint with Reitz) Every model of set theory is the eta-th inner mantle of a class forcing extension for any ordinal eta in the model. On the negative side: The sequence of inner mantles may fail to carry through at limit stages. Specifically, it is consistent that the omega-th inner mantle not be a definable class and it is consistent that it be a definable inner model of ¬AC.

 An important property of a Gromov hyperbolic space is that every path that is locally a quasi-geodesic is globally a quasi-geodesic. A theorem of Gromov states that this is a characterization of hyperbolicity, which means that all the properties of hyperbolic spaces and groups can be traced back to this simple fact. In this talk we generalize this property by considering only Morse quasi-geodesics.

We show that not only does this allow us to consider a much larger class of examples, such as CAT(0) spaces, hierarchically hyperbolic spaces and fundamental groups of 3-manifolds, but also we can effortlessly generalize several results from the theory of hyperbolic groups that were previously unknown in this generality.
 

Identifying discrete patterns in binary data is an important dimensionality reduction tool in machine learning and data mining. In this paper, we consider the problem of low-rank binary matrix factorisation (BMF) under Boolean arithmetic. Due to the NP-hardness of this problem, most previous attempts rely on heuristic techniques. We formulate the problem as a mixed integer linear program and use a large scale optimisation technique of column generation to solve it without the need of heuristic pattern mining. Our approach focuses on accuracy and on the provision of optimality guarantees. Experimental results on real world datasets demonstrate that our proposed method is effective at producing highly accurate factorisations and improves on the previously available best known results for 16 out of 24 problem instances.

--

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.

An algebra $R$ is said to satisfy the Dixmier-Moeglin equivalence if a prime ideal $P$ of $R$ is primitive if and only if it is rational, if and only if it is locally closed, and a commonly studied problem in non-commutative algebra is to classify rings satisfying this equivalence, e.g. $U(\mathfrak g)$ for a finite dimensional Lie algebra $\mathfrak g$. We explore methods of generalising this to a $p$-adic setting, where we need to weaken the statement. Specifically, if $\hat R$ is the $p$-adic completion of a $\mathbb Q_p$-algebra $R$, rather than approaching the Dixmier-Moeglin equivalence for $\hat R$ directly, we instead compare the classes of primitive, rational and locally closed prime ideals of $\hat R$ within suitable "deformations". The case we focus on is where $R=U(L)$ for a $\mathbb Z_p$-Lie algebra $L$, and the deformations have the form $\hat U(p^n L)$, and we aim to prove a version of the equivalence in the instance where $L$ is nilpotent.

In this talk, we will present an approach to constrained optimisation when the set of constraints is a smooth manifold. This setting is of particular interest in data science applications, as many interesting sets of matrices have a manifold structure. We will show how we may couple classic ideas from differential geometry with modern methods such as autodifferentiation to simplify optimisation problems from spaces with a difficult topology (e.g. problems with orthogonal or fixed-rank constraints) to problems on ℝⁿ where we can use any classical optimisation methods to solve them. We will also show how to use these methods to automatically compute quantities such as the Riemannian gradient and Hessian. We will present the library GeoTorch that allows for putting these kind of constraints within models written in PyTorch by adding just one line to the model. We will also comment on some convergence results if time allows.

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.

Multivariate point processes are used to model event-type data in a wide range of domains. One interesting application is to model the emission of electric impulses of biological neurons. In this context, the point process model needs to capture the time-dependencies and interactions between neurons, which can be of two kinds: exciting or inhibiting. Estimating these interactions, and in particular the functional connectivity of the neurons are problems that have gained a lot of attention recently. The general nonlinear Hawkes process is a powerful model for events occurring at multiple locations in interaction. Although there is an extensive literature on the analysis of the linear model, the probabilistic and statistical properties of the nonlinear model are still mainly unknown. In this paper, we consider nonlinear Hawkes models and, in a Bayesian nonparametric inference framework, derive concentration rates for the posterior distribution.  We also infer the graph of interactions between the dimensions of the process and prove that the posterior distribution is consistent on the graph adjacency matrix.

 In this talk I will discuss possible extensions of the euclidean quantitative isoperimetric inequality to compact Riemannian manifolds. 
This is joint work with O. Chodosh (Stanford) and M. Engelstein (University of Minnesota).