Past

Ductal carcinoma is one of the most common cancers among women, and the main cause of death is the formation of metastases. The development of metastases is caused by cancer cells that migrate from the primary tumour site (the mammary duct) through the blood vessels and extravasating they initiate metastasis. In my talk, I present a multi-compartment mathematical model which mimics the dynamics of tumoural cells in the mammary duct, in the circulatory system and in the bone. Using a branching process approach, the model describes the relation between the survival times and the four markers mainly involved in metastatic breast cancer (EPCAM, CD47, CD44 and MET). In particular, the model takes into account the gene expression profile of circulating tumour cells to predict personalised survival probability. Gene expression data of metastatic breast cancer have been used to validate the model. The administration of drugs as bisphosphonates is also included in order to analyse the dynamic changes induced by the therapy.

Stochastic and deterministic processes are merged together to describe cancer progression and obtain personalised survival analysis based on the gene expression levels of each patient. The main aim of the talk is showing that Mathematics can have a strong impact in speeding cancer research, predicting survival probability and selecting the best cancer treatment. 

tba

Parallelizing the time domain in numerical simulations is non-intuitive, but has been proven to be possible using various algorithms like parareal, PFASST and RIDC. Temporal parallelizations adds an entire new dimension to parallelize and significantly enhances use of super computing resources. Exploiting this technique serves as a big step towards exascale computation.

Starting with relatively simple problems, the parareal algorithm (Lions et al, A ''parareal'' in time discretization of PDE's, 2001) has been successfully applied to various complex simulations in the last few years (Samaddar et al, Parallelization in time of numerical simulations of fully-developed plasma turbulence using the parareal algorithm, 2010). The algorithm involves a predictor-corrector technique.

Numerical studies of the edge of magnetically confined, fusion plasma are an extremely challenging task. The complexity of the physics in this regime is particularly increased due to the presence of neutrals as well as the interaction of the plasma with the wall. These simulations are extremely computationally intensive but are key to rapidly achieving thermonuclear breakeven on ITER-like machines.

The SOLPS code package (Schneider et al, Plasma Edge Physics with B2‐Eirene, 2006) is widely used in the fusion community and has been used to design the ITER divertor. A reduction of the wallclock time for this code has been a long standing goal and recent studies have shown that a computational speed-up greater than 10 is possible for SOLPS (Samaddar et al, Greater than 10x Acceleration of fusion plasma edge simulations using the Parareal algorithm, 2014), which is highly significant for a code with this level of complexity.

In this project, the aim is to explore a variety of cases of relevance to ITER and thus involving more complex physics to study the feasibility of the algorithm. Since the success of the parareal algorithm heavily relies on choosing the optimum coarse solver as a predictor, the project will involve studying various options for this purpose. The tasks will also include performing scaling studies to optimize the use of computing resources yielding maximum possible computational gain.

Let M be an algebraic curve over an algebraically closed field and let
$(M, ...)$ be a strongly minimal non-locally modular structure with
basic relations definable in the full Zariski language on $M$. In this
talk I will present the proof of the fact that $(M, ...)$ interprets
an algebraically closed field.

I will introduce the profinite completion as a way of aggregating information about the finite-sheeted covers of a 3-manifold, and discuss the state of the homeomorphism problem for 3-manifolds in this context; in particular, for geometrizable 3-manifolds.

We argue that Credit Default Swap (CDS) premia for safe-haven sovereigns, like Germany and the United States, are driven to a large extent by regulatory requirements under which  derivatives dealing banks have an incentive to buy CDS to hedge counterparty credit risk of their counterparties.
We explain the mechanics of the regulatory requirements and develop a model in which derivatives dealers, who have a derivatives exposure with sovereigns, need CDS for capital relief. End users without exposure to the sovereigns sell the CDS and require a positive premium equivalent to the capital requirement. The model's predictions are confirmed using data on several sovereigns.

Joint with OMI

I will discuss some new results on the Iwasawa theory for the $3$-dimensional symmetric square Galois representation of a modular form, using the Euler system of Beilinson-Flach elements I constructed in joint work with Kings, Lei and Loeffler.

Networks form the backbones of a wide variety of complex systems,
ranging from food webs, gene regulation and social networks to
transportation networks and the internet. Due to the sheer size and
complexity of many of theses systems, it remains an open challenge to
formulate general descriptions of their large-scale structures.
Although many methods have been proposed to achieve this, many of them
yield diverging descriptions of the same network, making both the
comparison and understanding of their results very
difficult. Furthermore, very few methods attempt to gauge the
statistical significance of the uncovered structures, and hence the
majority cannot reliably separate actual structure from stochastic
fluctuations.  In this talk, I will show how these issues can be tackled
in a principled fashion by formulating appropriate generative models of
network structure that can have their parameters inferred from data. I
will also consider the comparison between a variety of generative
models, including different structural features such as degree
correction, where nodes with arbitrary degrees can belong to the same
group, and community overlap, where nodes are allowed to belong to more
than one group. Because such model variants possess an increased number
of parameters, they become prone to overfitting. We demonstrate how
model selection based on the minimum description length criterion and
posterior odds ratios can fully account for the increased degrees of
freedom of the larger models, and selects the most appropriate trade-off
between model complexity and quality of fit based on the statistical
evidence present in the data.

Throughout the talk I will illustrate the application of the methods
with many empirical networks such as the internet at the autonomous
systems level, the global airport network, the network of actors and
films, social networks, citations among websites, co-occurrence of
disease-causing genes and many others.
 

Firstly, we will discuss how the category of strict polynomial functors can be endowed with a monoidal structure, including concrete calculations. It is well-known that the above category is equivalent to the category of modules over the Schur algebra. The so-called Schur functor in turn relates the category of modules over the Schur algebra to the category of representations of the symmetric group which posseses a monoidal structure given by the Kronecker product. We show that the Schur functor is monoidal with respect to these structures.
Finally, we consider the right and left adjoints of the Schur functor. We explain how these can be expressed in terms of one another using Kuhn duality and the central role the monoidal structure on strict polynomial functors plays in this context.
 

The talk will introduce the concepts of multilevel optimization and motivate them in the context of problems arising from the discretization of infinite dimensional applications. It will be shown how optimization methods can accomodate a number of useful (and classical) ideas from the multigrid
community, and thereby produce substantial efficiency improvements compared to existing large-scale minimization techniques.  Two different classes of multilevel methods will be discussed: trust-region and linesearch algorithms.
The first class will be described in the context of a multilevel generalization of the well-known trust-region-Newton method.  The second will focus on limited-memory quasi-Newton algorithms.  Preliminary numerical results will be presented which indicate that both types of multilevel algorithms may be practically very advantageous.

We study a simple one-dimensional coupled heat wave system, obtaining a sharp estimate for the rate of energy decay of classical solutions. Our approach is based on the asymptotic theory of $C_0$-semigroups and in particular on a result due to Borichev and Tomilov (2010), which reduces the problem of estimating the rate of energy decay to finding a growth bound for the resolvent of the semigroup generator. This technique not only leads to an optimal result, it is also simpler than the methods used by other authors in similar situations and moreover extends to problems on higher-dimensional domains. Joint work with C.J.K. Batty (Oxford) and L. Paunonen (Tampere).

I will discuss a notoriously hard problem in group theory known as the flat closing conjecture. This states that a group with a finite classifying space is either hyperbolic or contains a Baumslag-Solitar Subgroup. I will give some strategies to try and create a counterexample to this conjecture. 

Quantum computers derive their computational power from the ability to manipulate superposition states of quantum registers. The generic quantum attack against a symmetric encryption scheme with key length n using Grover's algorithm has O(2^(n/2)) time complexity. For this kind of attack, an adversary only needs classical access to an encryption oracle. In this talk I discuss adversaries with quantum superposition access to encryption and decryption oracles. First I review and extend work by Kuwakado and Morii showing that a quantum computer with superposition access to an encryption oracle can break the Even-Mansour block cipher with key length n using only O(n) queries. Then, improving on recent work by Boneh and Zhandry, I discuss indistinguishability notions in chosen plaintext and chosen ciphertext attacks by a quantum adversary with superposition oracle access and give constructions that achieve these security notions.

Following an idea of Bridgeland, we study the operator on the K-group of algebraic stacks, which takes a stack to its inertia stack.  We prove that the inertia operator is diagonalizable when restricted to nice enough stacks, including those with algebra stabilizers.  We use these results to prove a structure theorem for the motivic Hall algebra of a projective variety, and give a more conceptual definition of virtually indecomposable stack function.  This is joint work with Pooya Ronagh.

We discuss a new setting of algorithmic problems in random graphs, studying the minimum number of queries one needs to ask about the adjacency between pairs of vertices of $G(n,p)$ in order to typically find a subgraph possessing a certain structure. More specifically, given a monotone property of graphs $P$, we consider $G(n,p)$ where $p$ is above the threshold probability for $P$ and look for adaptive algorithms which query significantly less than all pairs of vertices in order to reveal that the property $P$ holds with high probability. In this talk we focus particularly on the properties of containing a Hamilton cycle and containing paths of linear size. The talk is based on joint work with Asaf Ferber, Michael Krivelevich and Benny Sudakov.

Our work (which is joint with Simon Smith) began as a study of the structure of infinite permutation groups $G$ in which point stabilisers are finite and all infinite normal subgroups are transitive. That led to two variations. One is the generalisation in which point stabilisers are merely assumed to satisfy min-{\sc N}, the minimal condition on normal subgroups. The groups $G$ are then of two kinds. Either they have a maximal finite normal subgroup, modulo which they have either one or two minimal non-trivial normal subgroups, or they have a regular normal subgroup $M$ which is a divisible abelian $p$-group of finite rank. In the latter case the point stabilisers are finite and act irreducibly on the socle of~$M$. This leads to our second variation, which is a study of the finite linear groups that can arise.

The talk will describe a new "Brute Force Optimizer" whose objective is to provide a very versatile derivative-free Matlab package for bound-constrained optimization, with the distinctive feature that it can be trained to improve its own performance on classes of problems specified by the user (rather than on a single-but-wide problem class chosen by the algorithm developer).  In addition, BFO can be used to optimize the performance of other algorithms and provides facilities for mixed-integer and multilevel problems, including constrained equilibrium calculations.

Starting from Hilbert's 10th problem, I will explain how to characterise the set of integers by non-solubility of a set of polynomial equations and discuss related challenges. The methods needed are almost entirely elementary; ingredients from algebraic number theory will be explained as we go along. No knowledge of first-order logic is necessary.

Starting from Hilbert's 10th problem, I will explain how to characterise the set of integers by non-solubility of a set of polynomial equations and discuss related challenges. The methods needed are almost entirely elementary; ingredients from algebraic number theory will be explained as we go along. No knowledge of first-order logic is necessary.

We study the adaptive finite element approximation of the Dirichlet problem $-\Delta u = f$ with zero boundary values using newest vertex bisection. Our approach is based on the minimization of the corresponding Dirichlet energy. We show that the maximums strategy attains every energy level with a number of degrees of freedom, which is proportional to the optimal number. As a consequence we achieve instance optimality of the error. This is a joint work with Christian Kreuzer (Bochum) and Rob Stevenson (Amsterdam).

Abstract: A diffusion process on a Riemannian manifold whose generator is one half of the Laplacian is called a Brownian motion. The mean local time of Brownian motion on a hypersurface will be considered, as will the situation in which a Brownian motion is conditioned to arrive in a fixed submanifold at a fixed positive time. Doing so provides motivation for the remainder of the talk, in which a probabilistic formula for the integral of the heat kernel over a submanifold is proved and used to deduce lower bounds, an asymptotic relation and derivative estimates applicable to the conditioned process.

We study symplectic invariants of the open symplectic manifolds X
obtained by plumbing cotangent bundles of spheres according to a
plumbing tree. We prove that certain models for the Fukaya category F(X)
of closed exact Lagrangians in X and the wrapped Fukaya category W(X)
are related by Koszul duality. As an application, we give explicit
computations of symplectic cohomology essentially for all trees. This is
joint work with Tolga Etg\"u.

Abstract: The martingale optimal transport aims to optimally transfer a probability measure to another along the class of martingales. This problem is mainly motivated by the robust superhedging of exotic derivatives in financial mathematics, which turns out to be the corresponding Kantorovich dual. In this paper we consider the continuous-time martingale transport on the Skorokhod space of cadlag paths. Similar to the classical setting of optimal transport, we introduce different dual problems and establish the corresponding dualities by a crucial use of the S-topology and the dynamic programming principle. This is a joint work with Gaoyue Guo and Nizar Touzi.

By regarding gravity as the convolution of left and right Yang-Mills theories together with a spectator scalar field in the bi-adjoint representation, we derive in linearised approximation the gravitational symmetries of general covariance, p-form gauge invariance, local Lorentz invariance and local supersymmetry from the flat space Yang-Mills symmetries of local gauge invariance and global super-Poincare. As a concrete example we focus on the new-minimal (12+12) off-shell version of simple four-dimensional supergravity obtained by tensoring the off-shell Yang-Mills multiplets (4+4,NL =1)and(3+0,NR =0). 

Across much of the ablation region of the western Greenland Ice Sheet, hydro-fracture events related to supraglacial lake drainages rapidly deliver large volumes of meltwater to the bed of the ice sheet. We investigate what triggers the rapid drainage of a large supraglacial lake using a Network Inversion Filter (NIF) to invert a dense local network of GPS observations over three summers (2011-2013). The NIF is used to determine the spatiotemporal variability in ice sheet behavior (1) prior to lake drainage, and in response to (2) vertical hydro-fracture crack propagation and closure, (3) the opening of a horizontal cavity at the ice-sheet bed that accommodates the rapid injection of melt-water, and (4) extra basal slip due to enhanced lubrication. We find that the opening and propagation of each summer’s lake-draining hydro-fracture is preceded by a local stress perturbation associated with ice sheet uplift and enhanced slip above pre-drainage background velocities. We hypothesize that these precursors are associated with the introduction of meltwater to the bed through neighboring moulin systems.

The first half of the lecture will begin by reviewing what is known about the
neural representation of faces in the primate visual system. How does the
visual system represent the spatial structure of faces, facial identity and
expression? We then discuss how depression is associated with negative
cognitive biases in the recognition of facial expression, whereby depressed
people interpret facial expressions more negatively. The second half of the
lecture presents computer simulations aimed at understanding how these facial
representations may develop through visual experience. We show how neural
representations of expression are linked to particular spatial relationships
between facial features. Building on this, we show how the synaptic connections
in the model may be rewired by visual training to eliminate the negative
cognitive biases seen in depression.

Pengyu Wei's title: Ranking ForexMaster Players

Abstract:

In this talk I will introduce ForexMaster, a simulated foreign exchange trading platform, and how I rank players on this platform. Different methods are compared. In particular, I use random forest and a carefully chosen feature set, which includes not only traditional performance measures like Sharp ratio, but also estimates from the Plackett-Luce ranking model, which has not been used in the financial modelling yet. I show players selected by this method have satisfactory out-of-sample performance, and the Plackett-Luce model plays an important role.

Alissa Kleinnijenhuis title: Stress Testing the European Banking System: Exposure Risk & Overlapping Portfolio Risk
Abstract:
Current regulatory stress testing, as for example done by the EBA, BoE and the FED, is microprudential, non-systemic. These stress tests do not take into account systemic risk, even though the official aim of the stress test is the "test the resilience of the financial system as a whole, and the individual banks therein, to another crisis".
 Two papers are being developed that look at the interconnections between banks. One paper investigates the systemic risk in the European banking system due to interbank exposures, using EBA data. The other paper, looks at the trade-off between individual and systemic risk with overlapping portfolios. The above two "channels of contagion" for systemic risk can be incorporated in stress tests to include systemic components to the traditional non-systemic stress tests.

Blenders and food processors have been around for years.  However, detailed understanding of the fluid and particle dynamics going on with in the multi-phase flow of the processing chamber as well as the influence of variables such as the vessel geometry, blade geometry, speeds, surface properties etc., are not well understood.  SharkNinja would like Oxford Universities help in developing a model that can be used to gain insight into fluid dynamics within the food processing chamber with the goal being to develop a system that will produce better food processing performance as well as predict loading on food processing elements to enable data driven product design.

Many vacuum cleaners sold claim “no loss of suction” which is defined as having only a very small reduction in peak air power output over the life of the unit under normal operating conditions.  This is commonly achieved by having a high efficiency cyclonic separator combined with a filter which the user washes at regular intervals (typically every 3 months).  It has been observed that some vacuum cleaners show an increase in peak air watts output after a small amount of dust is deposited on the filter.  This effect is beneficial since it prolongs the time between filter washing.  SharkNinja are currently working on validating their theory as to why this occurs.  SharkNinja would like Oxford University’s help in developing a model that can be used to better understand this effect and provide insight towards optimizing future designs.

Although a very simple system from a construction standpoint, creating a drip coffee maker that can be produce a range of coffee sizes from a single cup to a multi-cup carafe presents unique problems.  Challenges within this system result from varying pressure heads on the inlet side, accurate measurement of relatively low flow rates, fluid motive force generated by boilers, and head above the boiler on the outlet side.  Getting all of these parameters right to deliver the proper strength, proper temp, and proper volume of coffee requires in depth understanding of the fluid dynamics involved in the system.  An ideal outcome from this work would be an adaptive model that enables a fluid system model to be created from building blocks.  This system model would include component models for tubing, boilers, flow meters, filters, pumps, check valves, and the like.

We consider the decision problem of determining whether an exponential
polynomial has a real zero.  This is motivated by reachability questions
for continuous-time linear dynamical systems, where exponential
polynomials naturally arise as solutions of linear differential equations.

The decidability of the Zero Problem is open in general and our results
concern restricted versions.  We show decidability of a bounded
variant---asking for a zero in a given bounded interval---subject to
Schanuel's conjecture.  In the unbounded case, we obtain partial
decidability results, using Baker's Theorem on linear forms in logarithms
as a key tool.  We show also that decidability of the Zero Problem in full
generality would entail powerful new effectiveness results concerning
Diophantine approximation of algebraic numbers.

This is joint work with Ventsislav Chonev and Joel Ouaknine.

Deformation K-theory was introduced by G. Carlsson and gives an interesting invariant of a group G encoding higher homotopy information about its representation spaces. Lawson proved a relation between this object and a homotopy theoretic analogue of the representation ring. This talk will not contain many details, instead I will outline some basic constructions and hopefully communicate the main ideas.
 

A well-known result of Landsberger and Meilijson says that efficient risk-sharing rules for univariate risks are characterized by a so-called comonotonicity condition. In this talk, I'll first discuss a multivariate extension of this result (joint work with R.-A. Dana and A. Galichon). Then I will discuss the restrictions (in the form of systems of nonlinear PDEs) efficient risk sharing imposes on individual consumption as a function of aggregate consumption. I'll finally give an identification result on how to recover preferences from the knowledge of the risk sharing (joint work with M. Aloqeili and I. Ekeland).

The Möbius function plays a central role in number theory; both the prime number theorem and the Riemann Hypothesis are naturally formulated in terms of the amount of cancellations one gets when summing the Möbius function. In a recent joint work with Maksym Radziwill we have shown that the sum of the Möbius function exhibits cancellation in "almost all intervals" of arbitrarily slowly increasing length. This goes beyond what was previously known conditionally on the Riemann Hypothesis. Our result holds in fact in much greater generality, and has several further applications, some of which I will discuss in the talk. For instance the general result implies that between a fixed number of consecutive squares there is always an integer composed of only "small" prime factors. This settles a conjecture on "smooth" or "friable" numbers and is related to the running time of Lenstra's factoring algorithm.

Noise limits are one of the major constraints when designing
aircraft engines.  Acoustic liners are fitted in almost all civilian
turbofan engine intakes, and are being considered for use elsewhere in a
bid to further reduce noise.  Despite this, models for acoustic liners
in flow have been rather poor until recently, with discrepancies of 10dB
or more.  This talk will show why, and what is being done to model them
better.  In the process, as well as mathematical modelling using
asymptotics, we will show that state of the art Computational
AeroAcoustics simulations leave a lot to be desired, particularly when
using optimized finite difference stencils.

The relationship between the so-called ancient (backwards) solutions to the Navier-Stokes equations in the space or in a half space and the global well-posedness of initial boundary value problems for these equations will be explained. If time permits I will sketch details of an equivalence theorem and a proof of smoothness properties of mild bounded ancient solutions in the half space, which is a joint work with Gregory Seregin

This talk will be an easy introduction to some CAT(0) geometry. Among other things, we'll see why centralizers in groups acting geometrically on CAT(0) spaces split (at least virtually). Time permitting, we'll see why having a geometric action on a CAT(0) space is not a quasi-isometry invariant.

The Learning with Errors (LWE) problem has become a central building block of modern cryptographic constructions. We will discuss hardness results for concrete instances of LWE. In particular, we discuss algorithms proposed in the literature and give the expected resources required to run them. We consider both generic instances of LWE as well as small secret variants. Since for several methods of solving LWE we require a lattice reduction step, we also review lattice reduction algorithms and propose a refined model for estimating their running times. We also give concrete estimates for various families of LWE instances, provide a Sage module for computing these estimates and highlight gaps in the knowledge about algorithms for solving the Learning with Errors problem.

We investigate equivalence relations for quadratic forms that can be expressed in terms of algebro-geometric properties of their associated quadrics, more precisely, birational, stably birational and motivic equivalence, and isomorphism of quadrics. We provide some examples and counterexamples and highlight some important open problems.

Igusa's p-adic zeta function $Z(s)$ attached to a polynomial $f$ in $N$ variables is a meromorphic function on the complex plane that encodes the numbers of solutions of the equation $f=0$ modulo powers of a prime $p$. It is expressed as a $p$-adic integral, and Igusa proved that it is rational in $p^{-s}$ using resolution of singularities and the change of variables formula. From this computation it is immediately clear that the order of a pole of $Z(s)$ is at most $N$, the number of variables in $f$. In 1999, Wim Veys conjectured that the only possible pole of order $N$ of the so-called topological zeta function of $f$ is minus the log canonical threshold of $f$. I will explain a proof of this conjecture, which also applies to the $p$-adic and motivic zeta functions. The proof is inspired by non-archimedean geometry and Mirror Symmetry, but the main technique that is used is the Minimal Model program in birational geometry. This talk is based on joint work with Chenyang Xu.

Modelling two-phase, incompressible flow with level set or volume-of-fluid formulations results in a variable coefficient Navier-Stokes system that is challenging to solve computationally. In this talk I will present work from a recent InFoMM CDT mini-project which looked to adapt current preconditioners for one-phase Navier-Stokes flows. In particular we consider systems arising from the application of finite element methodology and preconditioners which are based on approximate block factorisations. A crucial ingredient is a good approximation of the Schur complement arising in the factorisation which can be computed efficiently.

The Erdős–Ko–Rado theorem is a central result in extremal set theory which tells us how large uniform intersecting families can be. In this talk, I shall discuss some recent results concerning the 'stability' of this result. One possible formulation of the Erdős–Ko–Rado theorem is the following: if $n \ge 2r$, then the size of the largest independent set of the Kneser graph $K(n,r)$ is $\binom{n-1}{r-1}$, where $K(n,r)$ is the graph on the family of $r$-element subsets of $\{1,\dots,n\}$ in which two sets are adjacent if and only if they are disjoint. The following will be the question of interest. Delete the edges of the Kneser graph with some probability, independently of each other: is the independence number of this random graph equal to the independence number of the Kneser graph itself? I shall discuss an affirmative answer to this question in a few different regimes. Joint work with Bollobás and Raigorodskii, and Balogh and Bollobás.

The question that the Crepant Resolution Conjecture (CRC) wants to address is: given an orbifold X that admits a repant resolution Y, can we systematically compare the Gromov-Witten theories of the two spaces? That this should happen was first observed by physicists and the question was imported into mathematics by Y.Ruan, who posited it as the search for an isomorphism in the quantum cohomologies of the two spaces. In the last fifteen years this question has evolved and found different formulations which various degree of generality and validity. Perhaps the most powerful approach to the CRC is through Givental's formalism. In this case, Coates, Corti, Iritani and Tseng propose that the CRC should consist of the natural comparison of geometric objects constructed from the GW potential fo the space. We explore this approach in the setting of open GW invariants. We formulate an open version of the CRC using this formalism, and make some verifications. Our approach is well tuned with Iritani's approach to the CRC via integral structures, and it seems to suggest that open invariants should play a prominent role in mirror symmetry. 

Wave scattering problems arise in numerous applications in acoustics, electromagnetics and linear elasticity. In the boundary element method (BEM) one reformulates the scattering problem as an integral equation on the scatterer boundary, e.g. using Green’s identities, and then seeks an approximate solution of the boundary integral equation (BIE) from some finite-dimensional approximation space. The conventional choice is a space of piecewise polynomials; however, in the “high frequency” regime when the wavelength is small compared to the size of the scatterer, it is computationally expensive to resolve the highly oscillatory wave solution in this way. The hybrid numerical-asymptotic (HNA) approach aims to reduce the computational cost by enriching the BEM approximation space with oscillatory functions, carefully chosen to capture the high frequency asymptotic solution behaviour. To date, the HNA methodology has been implemented almost exclusively in a Galerkin variational framework. This has many attractive features, not least the possibility of proving rigorous convergence results, but has the disadvantage of requiring numerical evaluation of high dimensional oscillatory integrals. In this talk I will present the results of some investigations carried out with my MSc student Emile Parolin into collocation-based implementations, which involve lower-dimensional integrals, but appear harder to analyse in terms of convergence and stability.