We’ll have an open discussion about the ways in which Mathematics is very euro-centric and how we can act, as students and educators, to change this.

Recent years have witnessed a surge of interest in deep learning methods to tackle inverse problems arising in various domains such as medical imaging, remote sensing, and the arts and humanities. This talk offers an overview of recent advances in the foundations and applications of deep learning for inverse problems, with a focus on model-based deep learning methods. Concretely, this talk will overview our work relating to theoretical advances in the area of mode-based learning, including learning guarantees; algorithmic advances in model-based learning; and, finally it will showcase a portfolio of emerging signal & image processing challenges that benefit from model based learning, including image separation / deconvolution challenges arising in the arts and humanities.

Bio:

Miguel Rodrigues is a Professor of Information Theory and Processing at University College London; he leads the Information, Inference and Machine Learning Lab at UCL, and he has also been the founder and director of the master programme in Integrated Machine Learning Systems at UCL. He has also been the UCL Turing University Lead and a Turing Fellow with the Alan Turing Institute — the UK National Institute of Data Science and Artificial Intelligence.

He held various appointments with various institutions worldwide including Cambridge University, Princeton University, Duke University, and the University of Porto, Portugal. He obtained the undergraduate degree in Electrical and Computer Engineering from the Faculty of Engineering of the University of Porto, Portugal and the PhD degree in Electronic and Electrical Engineering from University College London.

Dr. Rodrigues's research lies in the general areas of information theory, information processing, and machine learning. His most relevant contributions have ranged from the information-theoretic analysis and design of communications systems, information-theoretic security, information-theoretic analysis and design of sensing systems, and the information-theoretic foundations of machine learning.

He serves or has served as Editor of IEEE BITS, Editor of the IEEE Transactions on Information Theory, and Lead Guest Editor of various Special Issues of the IEEE Journal on Selected Topics in Signal Processing, Information and Inference, and Foundations and Trends in Signal Processing.

Dr. Rodrigues has been the recipient of various prizes and awards including the Prize for Merit from the University of Porto, the Prize Engenheiro Cristian Spratley, the Prize Engenheiro Antonio de Almeida, fellowships from the Portuguese Foundation for Science and Technology, and fellowships from the Foundation Calouste Gulbenkian. Dr. Rodrigues research on information-theoretic security has also attracted the IEEE Communications and Information Theory Societies Joint Paper Award 2011.  

He has also been elevated to Fellow of the Institute of Electronics and Electrical Engineers (IEEE) for his contributions to the ‘multi-modal data processing and reliable and secure communications.’

Systems of high-contrast resonators can be used to control and manipulate wave-matter interactions at scales that are much smaller than the operating wavelengths. The aim of this talk is to review recent studies of ordered and disordered systems of subwavelength resonators and to explain some of their topologically protected localization properties. Both reciprocal and non-reciprocal systems will be considered.
 

The mod-$p$ cohomology of uniform pro-$p$ groups has been calculated by Lazard in the 1960s. Motivated by recent considerations in the mod-$p$ Langlands program, we consider the problem of extending his results to the case of compact $p$-adic Lie groups $G$ that are $p$-saturable but not necessarily uniform pro-$p$: when $F$ is a finite extension of $\mathbb{Q}_p$ and $p$ is sufficiently large, this class of groups includes the so-called pro-$p$ Iwahori subgroups of $SL_n(F)$. In general, there is a spectral sequence due to Serre and Lazard that relates the mod-$p$ cohomology of $G$ to the cohomology of its associated graded mod-$p$ Lie algebra $\mathfrak{g}$. We will discuss certain sufficient conditions on $p$ and $G$ that ensure that this spectral sequence collapses. When these conditions hold, it follows that the mod-$p$ cohomology of $G$ is isomorphic to the cohomology of the Lie algebra $\mathfrak{g}$.

Fundamental to the study of hyperbolic groups is their Gromov boundaries. The classical Cannon--Thurston map for a closed fibered hyperbolic 3-manifolds relates two such boundaries: it gives a continuous surjection from the boundary of the surface group (a circle) to the boundary of the 3-manifold group (a 2-sphere). Mj (Mitra) generalized this to all hyperbolic groups with hyperbolic normal subgroups. A generalization of the Gromov boundary to all finitely generated groups is called the Morse boundary. It collects all the "hyperbolic-like" rays in a group. In this talk we will discuss Cannon--Thurston maps for Morse boundaries. This is joint work with Ruth Charney, Antoine Goldsborough, Alessandro Sisto and Stefanie Zbinden.

I will review recent progress on extending the Minimal Model Program to algebraically integrable foliations, focusing on applications such as the canonical bundle formula and recent results toward the boundedness of Fano foliations.

Guentner, Willet and Yu defined a notion of dynamic asymptotic dimension for an étale groupoid that can be used to bound the nuclear dimension of its groupoid C*-algebra.  To have finite dynamic asymptotic dimension, the isotropy subgroups of the groupoid must be locally finite.  I will discuss 1) how to use similar ideas to bound the nuclear dimension of the C*-algebra of a groupoid with `large' isotropy subgroups and 2) the limitations of that approach. In an application to the C*-algebra of a directed graph,  if the C*-algebra is stably finite, then its nuclear dimension is at most 1.  This is joint work with Dana Williams. 

TBA

TBA

Many matrices that arise in scientific computing and in data science have internal structure that can be exploited to accelerate computations. The focus in this talk will be on matrices that are either of low rank, or can be tessellated into a collection of subblocks that are either of low rank or are of small size. We will describe how matrices of this nature arise in the context of fast algorithms for solving PDEs and integral equations, and also in handling "kernel matrices" from computational statistics. A particular focus will be on randomized algorithms for obtaining data sparse representations of such matrices.

At the end of the talk, we will explore an unorthodox technique for discretizing elliptic PDEs that was designed specifically to play well with fast algorithms for dense structured matrices.

The question can be formulated as a statistical hypothesis asserting that the distribution of the shapes of closed curves representing outlines of cell nuclei in a spatial domain is independent of the distribution of their locations. The key challenge in developing a procedure to test the hypothesis from a sample of spatially indexed curves (e.g. from an image) lies in how symmetries in the data are accounted for: shape of a curve is a property that is invariant to similarity transformations and reparameterization, and the shape space is thus an infinite-dimensional quotient space. Starting with a convenient geometry for the shape space developed over the last few years, I will discuss dependence measures and their estimates for spatial point processes with shape-valued marks, and demonstrate their use in testing for spatial independence of marks in a breast cancer application.  

TBD

I will present some contributions to the quasiisometry classification of solvable Lie groups of exponential growth that we obtain using sublinear bilipschitz equivalences, which are generalized quasiisometries. This is joint work with Ido Grayevsky.

The landmark completion of the Elliott classification program for unital separable simple nuclear C*-algebras saw three regularity properties rise to prominence: Z-stability, a C*-algebraic analogue of von Neumann algebras' McDuffness; finite nuclear dimension, an operator algebraic version of having finite Lebesgue dimension; and strict comparison, a generalization of tracial comparison in II_1 factors. Given their relevance to classification, most of the investigations into their interplay have focused on the simple nuclear case.

 The purpose of this talk is to advertise the general study of these properties and discuss their applications both within and outside operator algebras. Concretely, I will explain how characterizing when certain twisted group C*-algebras are Z-stable can provide new partial solutions to a well-known problem in generalized time-frequency analysis; this is joint work with U. Enstad. If time allows, I will also briefly discuss how a different incarnation of tracial comparison (finite radius of comparison) for non-commutative tori relates to the existence of smooth Gabor frames; this last part is joint work with U. Enstad and also H. Thiel.

TBA

A fundamental question in neuroscience is to understand how information is represented in the activity of  tens of thousands of neurons in the brain. Towards this end, low-rank matrix and tensor decompositions are commonly used to identify correlates of behavior in high-dimensional neural data. In this talk I will first present a novel tensor decomposition based on the slice rank which is able to disentangle mixed modes of covarying patterns in data tensors. Second, to compliment this statistical approach, I will present our recent dynamical systems modelling of neural activity over learning. Rather than factorizing data tensors themselves, we instead fit a dynamical system to the data, while constraining the tensor of parameters to be low rank. Together these projects highlight how applications in neural data can inspire new classes of low-rank models.

TBA

Sard’s theorem asserts that the set of critical values of a smooth map from one Euclidean space to another one has measure zero. A version of this result for infinite-dimensional Banach manifolds was proven by Smale for maps with Fredholm differential. However, when the domain is infinite dimensional and the range is finite dimensional, the result is not true – even under the assumption that the map is “polynomial” – and a general theory is still lacking. In this seminar, I will provide sharp quantitative criteria for the validity of Sard’s theorem in this setting, obtained combining a functional analysis approach with new tools in semialgebraic geometry. As an application, I will present new results on the Sard conjecture in sub-Riemannian geometry. Based on a joint work with A. Lerario and L. Rizzi.

Parameters in mathematical models are often impossible to determine fully or accurately, and are hence subject to uncertainty. By modelling the input parameters as stochastic processes, it is possible to quantify the uncertainty in the model outputs. 

In this talk, we employ the multilevel Monte Carlo (MLMC) method to compute expected values of quantities of interest related to partial differential equations with random coefficients. We make use of the circulant embedding method for sampling from the coefficient, and to further improve the computational complexity of the MLMC estimator, we devise and implement the smoothing technique integrated into the circulant embedding method. This allows to choose the coarsest mesh on the  first level of MLMC independently of the correlation length of the covariance function of the random  field, leading to considerable savings in computational cost.

 

Please note; this talk is hosted by Rutherford Appleton Laboratory, Harwell Campus, Didcot, OX11 0QX

TBA