Many theorems and conjectures in prime number theory are equivalent to finding solutions to certain linear equations in primes -- witness Goldbach's conjecture, the twin prime conjecture, Vinogradov's theorem, finding k-term arithmetic progressions, etcetera. Classically these problems were attacked using Fourier analysis -- the 'circle' method -- which yielded some success, provided that the number of variables was sufficiently large. More recently, a long research programme of Ben Green and Terence Tao introduced two deep and wide-ranging techniques -- so-called 'higher order Fourier analysis' and the 'transference principle' -- which reduces the number of required variables dramatically. In particular, these methods give an asymptotic formula for the number of k-term arithmetic progressions of primes up to X. In this talk we will give a brief survey of these techniques, and describe new work of the speaker, partially ongoing, which applies the Green-Tao machinery to count prime solutions to certain linear inequalities in primes -- a 'higher order Davenport-Heilbronn method'. 

A question by Poonen asks whether iterating the étale-Brauer set can give a finer obstruction set. We tackle the algebraic version of Poonen's question and give, in many cases, a negative answer.

On the railway network, for example, there is a large base of installed equipment with a useful life of many years.  This equipment has condition monitoring that can flag a fault when a measured parameter goes outside the permitted range.  If we can use existing measurements to predict when this would occur, preventative maintenance could be targeted more effectively and faults reduced.  As an example, we will consider the current supplied to a points motor as a function of time in each operational cycle.

“Market-Basket (MB) and Household (HH) data provide a fertile substrate for the inference of association between marketing activity (e.g.: prices, promotions, advertisement, etc.) and customer behaviour (e.g.: customers driven to a store, specific product purchases, joint product purchases, etc.). The main aspect of MB and HH data which makes them suitable for this type of inference is the large number of variables of interest they contain at a granularity that is fit for purpose (e.g.: which items are bought together, at what frequency are items bought by a specific household, etc.).

A large number of methods are available to researchers and practitioners to infer meaningful networks of associations between variables of interest (e.g.: Bayesian networks, association rules, etc.). Inferred associations arise from applying statistical inference to the data. In order to use statistical association (correlation) to support an inference of causal association (“which is driving which”), an explicit theory of causality is needed.

Such a theory of causality can be used to design experiments and analyse the resultant data; in such a context certain statistical associations can be interpreted as evidence of causal associations.

On observational data (as opposed to experimental), the link between statistical and causal associations is less straightforward and it requires a theory of causality which is formal enough to support an appropriate calculus (e.g.: do-calculus) of counterfactuals and networks of causation.

My talk will be focused on providing retail analytic problems which may motivate an interest in exploring causal calculi’s potential benefits and challenges.”

To face the advent of multicore processors and the ever increasing complexity of hardware architectures, programming
models based on DAG parallelism regained popularity in the high performance, scientific computing community. Modern runtime systems offer a programming interface that complies with this paradigm and powerful engines for scheduling the tasks into which the application is decomposed. These tools have already proved their effectiveness on a number of dense linear algebra applications. 

In this talk we present the design of task-based sparse direct solvers on top of runtime systems. In the context of the
qr_mumps solver, we prove the usability and effectiveness of our approach with the implementation of a sparse matrix multifrontal factorization based on a Sequential Task flow parallel programming model. Using this programming model, we developed features such as the integration of dense 2D Communication Avoiding algorithms in the multifrontal method allowing for better scalability compared to the original approach used in qr_mumps.

Following this approach, we move to heterogeneous architectures where task granularity and scheduling strategies are critical to achieve performance. We present, for the multifrontal method, a hierarchical strategy for data partitioning and a scheduling algorithm capable of handling the heterogeneity of resources.   Finally we introduce a memory-aware algorithm to control the memory behavior of our solver and show, in the context of multicore architectures, an important reduction of the memory footprint for the multifrontal QR factorization with a small impact on performance.