June 13th Lightning Speakers
Flexible and scalable modelling of event sequences, by Professor Ioannis Kosmidis
In this lightning talk, Professor Kosmidis will present recent collaborative work on a general framework for defining flexible point process models and associated learning algorithms to model sequences of events.
He will focus on applying the framework to modelling in-game event sequences from the English Premier League, drawing inferences about previously unquantified characteristics of the game dynamics and extracting event-specific team abilities and rankings.
Distributable computing with sequential Monte Carlo by Professor Adam Johansen
About: Professor Johansen's interests include many areas of computational statistics, broadly interpreted.
In this lightning talk, he will touch on a few areas in which a non-standard variant of a simulation-based method popular in engineering and computational statistics can be employed to tackle difficult estimation problems in a way amenable to distributed implementation.
Statistics data science in science and engineering problems by Dr Julia Brettschneider
About: This talk will highlight central challenges in using data science methods to contribute to the solution of problems in science and engineering from a statistical perspective.
In particular, Dr Brettschneider will demonstrate the use of spatial statistics (e.g. in checking the state of digital X-ray detectors and in microscopic imaging) and highlight pillars of data quality.
Gamma-convergence of Onsager--Machlup functionals and MAP estimation in non-parametric Bayesian inverse problems by Tim Sullivan
The Bayesian solution to a statistical inverse problem can be summarised by a mode of the posterior distribution, i.e. a MAP estimator. The MAP estimator essentially coincides with the (regularised) variational solution to the inverse problem, seen as minimisation of the Onsager--Machlup functional of the posterior measure. An open problem in the stability analysis of inverse problems is to establish a relationship between the convergence properties of solutions obtained by the variational approach and by the Bayesian approach. To address this problem, we propose a general convergence theory for modes that is based on the $\Gamma$-convergence of Onsager--Machlup functionals, and apply this theory to Bayesian inverse problems with Gaussian and edge-preserving Besov priors.
Location, location, location: Managing and analysing large spatio-temporal data streams by Professor Hakan Ferhatosmanoglu
With the prevalence of positioning devices and mobile services, massive amounts of location sequences are being generated continuously. Querying and learning over small-sized representations of trajectory data are needed for a wide variety of applications, such as monitoring road networks and disease spread.
In this talk, Hakan will discuss his work on managing and analysing spatio-temporal data streams, from privacy preserving collection to query-friendly compression of such data. The talk will include a near real-time traffic monitoring system built in collaboration with Transport for West Midlands.
