Past

We propose a powerful prediction algorithm built upon Gaussian
processes (GPs). They are particularly useful for their flexibility,
facilitating accurate prediction even in the absence of strong physical models. GPs further allow us to work within a completely Bayesian framework. As such, we show how the hyperparameters of our system can be marginalised by use of Bayesian Monte Carlo, a principled method of approximate integration. We employ the error bars of the GP's prediction as a means to select only the most informative observations to store. This allows us to introduce an iterative formulation of the GP to give a dynamic, on-line algorithm. We also show how our error bars can be used to perform active data selection, allowing the GP to select where and when it should next take a measurement.

We demonstrate how our methods can be applied to multi-sensor prediction problems where data may be missing, delayed and/or correlated. In particular, we present a real network of weather sensors as a testbed for our algorithm.

The study of representation growth of infinite groups asks how the

numbers of (suitable equivalence classes of) irreducible n-dimensional

representations of a given group behave as n tends to infinity. Recent

works in this young subject area have exhibited interesting arithmetic

and analytical properties of these sequences, often in the context of

semi-simple arithmetic groups.

In my talk I will present results on the representation growth of some

classes of finitely generated nilpotent groups. They draw on methods

from the theory of zeta functions of groups, the (Kirillov-Howe)

coadjoint orbit formalism for nilpotent groups, and the combinatorics

of (finite) Coxeter groups.

The Kerr solutions to Einstein's equations describe rotating black holes. For the wave equation in flat-space and outside the non-rotating, Schwarzschild black holes, one method for proving decay is the vector-field method, which uses the energy-momentum tensor and vector-fields. Outside the Schwarzschild black hole, a key intermediate step in proving decay involved proving a Morawetz estimate using a vector-field which pointed away from the photon sphere, where null geodesics orbit the black hole. Outside the Kerr black hole, the photon orbits have a more complicated structure. By using the hidden symmetry of Kerr, we can replace the Morawetz vector-field by a fifth-order operator which, in an appropriate sense, points away from the photon orbits. This allows us to prove the necessary Morawetz estimate. From this we can prove a decay estimate of almost $t^{-1}$ for fixed $r$ and the corresponding decay rates at the event horizon and null infinity. The major innovation in this result is that, by using the hidden symmetries with the energy-momentum, we can avoid taking Fourier tranforms in time.

This is joint work with Lars Andersson.

This work in collaboration with J. Casado-Díaz deals with the asymptotic behaviour of two-dimensional linear conduction problems for which the sequence of conductivity matrices is bounded from below but not necessarily from above. On the one hand, we prove an extension in dimension two of the classical div-curl lemma, which allows us to derive a H-convergence type result for any L¹-bounded sequence of conductivity matrices. On the other hand, we obtain a uniform convergence result satisfied by the minimisers of a sequence of two-dimensional diffusion energies. This implies the closure for the L²-strong topology of $\Gamma$-convergence of the sets of equicoercive diffusion energies without assuming any bound from above. A few counter-examples in dimension three, connected with the appearance of non-local effects, show the specificity of dimension two in the two previous compactness results.

Can investors with irrational beliefs be neglected as long as they are rational on average ? Does unbiased disagreement lead to trades that cancel out with no consequences on prices, as implicitly assumed by the traditional models ? We show in this paper that there is an important impact of unbiased disagreement on the behavior of financial markets, even though the pricing formulas are "on average" (over the states of the world) unchanged. In particular we obtain time varying, mean reverting and countercyclical (instead of constant in the standard model) market prices of risk, mean reverting and procyclical (instead of constant) risk free rates, decreasing (instead of flat) yield curves in the long run, possibly higher returns and higher risk premia in the long run (instead of a flat structure), momentum in stock returns in the short run, more variance on the state price density, time and state varying (instead of constant) risk sharing rules, as well as more important and procyclical trading volumes. These features seem consistent with the actual (or desirable) behavior of financial markets and only result from the introduction of unbiased disagreement.

Inverse problems are often ill-posed, with solutions that depend sensitively on data. Regularization of some form is often used to counteract this. I will describe an approach to regularization, based on a Bayesian formulation of the problem, which leads to a notion of well-posedness for inverse problems, at the level of probability measures.

The stability which results from this well-posedness may be used as the basis for understanding approximation of inverse problems in finite dimensional spaces. I will describe a theory which carries out this program.

The ideas will be illustrated with the classical inverse problem for the heat equation, and then applied to so more complicated inverse problems arising in data assimilation, such as determining the initial condition for the Navier-Stokes equation from observations.

I will present a new proof of Mumford's conjecture on the rational cohomology of moduli spaces of curves, which is substantially different from those given by Madsen--Weiss and Galatius--Madsen--Tillmann--Weiss: in particular, it makes no use of Harer--Ivanov stability for the homology of mapping class groups, which played a decisive role in the previously known proofs. This talk represents joint work with Soren Galatius.

Presheaves on categories crop up everywhere! In this talk, I'll give a

gentle introduction to 2-categories, and discuss the notion of a

presheaf on a 2-category. In particular, we'll consider which

2-categories such a presheaf might take values in. Only a little

familiarity with the notion of a category will be assumed!

Triply graded link homology (introduced by Khovanov and Rozansky) is a

categorification of the HOMFLYPT polynomial. In this talk I will discuss

recent joint work with Ben Webster which gives a geometric construction of this invariant in terms of equivariant constructible sheaves. In this

framework the Reidemeister moves have quite natural geometric proofs. A

generalisation of this construction yields a categorification of the

coloured HOMFLYPT polynomial, constructed (conjecturally) by Mackay, Stosic and Vaz. I will also describe how this approach leads to a natural formula for the Jones-Ocneanu trace in terms of the intersection cohomology of Schubert varieties in the special linear group.

Numerical computer codes implementing physics based models are the backbone of today's mechanical/aerospace engineering analysis and design methods. Such computational codes can be extremely expensive consisting of several millions of degrees of freedom. However, large models even with very detailed physics are often not enough to produce credible numerical results because of several types of uncertainties which exist in the whole process of physics based computational predictions. Such uncertainties include, but not limited to (a) parametric uncertainty (b) model inadequacy; (c) uncertain model calibration error coming from experiments and (d) computational uncertainty. These uncertainties must be assessed and systematically managed for credible computational predictions. This lecture will discuss a random matrix approach for addressing these issues in the context of complex structural dynamic systems. An asymptotic method based on eigenvalues and eigenvectors of Wishart random matrices will be discussed. Computational predictions will be validated against laboratory based experimental results.

The parabolic Anderson model is the Cauchy problem for the heat equation with random potential.  It offers a case study for the possible effects that a random, or irregular environment can have on a diffusion process.  In this talk I review results obtained for an extreme case of heavy-tailed potentials, among the effects we discuss our intermittency, strong localisation and ageing.

A simple, yet efficient, model for data acquisition is to measure signals $x$ linearly through the action of a measurement matrix $A$ of size $n\times N$, namely $b = Ax$. Surprisingly, only $n

Let G be a compact semisimple Lie group. A classical paper of Atiyah and Bott (from 1982) studies the moduli space of flat G-bundles on a fixed Riemann surface S. Their approach completely determines the integral homology of this moduli space, using Morse theoretic methods. In the case where G is U(n), this moduli space is homotopy equivalent to the moduli space of holomorphic vector bundles on S which are "semi-stable". Previous work of Harder and Narasimhan determined the Betti numbers of this moduli space using the Weil conjectures. 20 years later, a Madsen and Weiss determined the homology of the moduli space of Riemann surfaces, in the limit where the genus of the surface goes to infinity.

My talk will combine these two spaces: I will describe the homology of the moduli space of Riemann surfaces S, equipped with a flat G-bundle E -> S, where we allow both the flat bundle and the surface to vary. I will start by reviewing parts of the Atiyah-Bott and Madsen-Weiss papers. Our main theorem will then be a rather easy consequence. This is joint work with Nitu Kitchloo and Ralph Cohen.

In this lecture, I will illustrate the cobordism hypothesis by presenting some examples. Exact content to be determined, depending on the interests of the audience.

In this lecture, I will give a more precise statement of the Baez-Dolan cobordism hypothesis, which gives a description of framed bordism (higher) categories by a universal mapping property. I'll also describe some generalizations of the cobordism hypothesis, which take into account the structure of diffeomorphism groups of manifolds and which apply to manifolds which are not necessarily framed.

In this lecture, I'll give an overview of some ideas from higher category theory which are needed to make sense of the Baez-Dolan cobordism hypothesis. If time permits, I'll present Rezk's theory of complete Segal spaces (a model for the theory of higher categories in which most morphisms are assumed to be invertible) and explain how bordism categories can be realized in this framework.

In this lecture, I will review Atiyah's definition of a topological quantum field theory. I'll then sketch the definition of a more elaborate structure, called an "extended topological quantum field theory", and describe a conjecture of Baez and Dolan which gives a classification of these extended theories.

PROBLEM STATEMENT:

Consider a set of measurements made by many sensors placed in a noisy environment, the noise is both temporally and spatially correlated and has time varying statistics. Given this environment, characterised by spatial and temporal scales of correlation, the challenge is to detect the presence of a weak, stationary signal described by smaller scales of temporal and spatial correlation.

Many current and future challenges involve detection of signals in the presence of other, similar, signals. The signal environment is extremely busy and thus the traditional process of detection of a signal buried in noise at reducing signal to noise ratio is no longer sufficient. Signals of interest may be at high SNR but need to be detected, classified, isolated and analysed as close to real time as is possible. All interfering signals are potentially signals of interest and all overlap in time and frequency.

Can the performance of signal detection algorithms be parameterised by some characteristic(s) of the signal environment?

A problem exists to detect and classify multiple signal types, but with a very low duty cycle for the receiver. In certain circumstances, very short windows of opportunity exist where the local signal environment can be sampled and the duty cycle of observation opportunities can be as low as 10%. The signals to be detected may be continuous or intermittent (burst) transmissions. Within these short windows, it is desirable to detect and classify multiple transmissions in terms of signal type (e.g. analogue or digital comms, navigation etc.) and location of transmitters. The low duty cycle of observations for the receiver makes this a challenging prospect.

Again, can the performance of signal detection algorithms be parameterised by some characteristic(s) of the signal environment?

The classic coating-flow problem first studied experimentally by Moffat and asymptotically by Pukhnachov in 1977 is reconsidered in the framework of multiple-timescale asymptotics. Two-timescale approximations of the height of the thin film coating a rotating horizontal circular cylinder are obtained from an asymptotic analysis, in terms of small gravitational and capillary parameters, of Pukhnachov's nonlinear evolution for the film thickness. The transition, as capillary effects are reduced, from smooth to shock-like solutions is examined, and interesting large-time dynamics in this case are determined via a multiple-timescale analysis of a Kuramoto-Sivashinsky equation. A pseudo-three-timescale method is proposed and demonstrated to improve the accuracy of the smooth solutions, and an asymptotic analysis of a modified Pukhnachov's equation, one augmented by inertial terms, leads to an expression for the critical Reynolds number above which the steady states first analysed by Moffatt and Pukhnachov cannot be realised. As part of this analysis, some interesting implications of the effects of different scalings on inertial terms is discussed. All theoretical results are validated by either spectral or extrapolated numerics.