Prior to the PhD defence by Changqing Lu on the 27th of February, 2025, of the thesis “Statistical and machine learning contributions to spatial and spatio-temporal point process modelling, with an application to Dutch fire risk prediction”, there will be a symposium with both national and international speakers. Talks will be given by Rémi Bardenet (University of Lille), Geurt Jongbloed (Delft University of Technology), Rasmus Waagepetersen (Aalborg University) and Ron de Wit (Twente Fire Brigade).
The topics of the talks include advanced methods for approximating smooth functions using determinantal designs, recent developments in addressing classical problems from stereology, and insights into the asymptotic normality of conditionally centered space-time processes. The program also explores data-driven approaches to risk management, with a focus on present and future applications for fire services.
09:15 – 09:30 | Welcome and opening (Carre 2N) |
09:30 – 09:55 | Rémi Bardenet: Approximating smooth functions with determinantal designs Abstract: We are interested in the approximation of a square-integrable function from a finite number of evaluations on a random set of nodes. We assume that the target function is smooth, in the sense that it belongs to a reproducing kernel Hilbert space (RKHS). This covers many classical cases of interest in signal processing, such as band-limited functions. I will present our results on fast convergence rates for the mean-square approximation error, in both RKHS- and $L^2$-norm, when the set of nodes is drawn from a suitable determinantal point process, or a mixture thereof. |
09:55 – 10:20 | Geurt Jongbloed: Recent developments in a classical problem from stereology Abstract: Wicksell’s problem was posed exactly a century ago in 1925 as “the corpuscle problem” by Sven Wicksell in Biometrika. It is a model where spheres of (iid) random sizes are distributed within a 3D opaque medium according to a homogeneous Poisson Process. Cutting the medium in two by a plane, reveals circular profiles of random sizes. The main problem is then to obtain the distribution of the sphere sizes, being only able to see the circle radii. The problem is an example of a stereological problem. Statistically, it is an inverse problem. Wicksell already derived the distribution of the observable circle radii in terms of that of the underlying sphere radii. There is also an explicit inverse transformation, expressing the distribution of interest in terms of that of the observables. In this presentation, we will present recently found adaptivity properties of the so-called isotonic inverse estimator in the classical problem, where the 3D particles are spheres. We will also introduce a more general formulation of the model, where the particles are more general convex bodies (all of the same shape, with random sizes). That model is also an inverse problem, but the transformations cannot be analytically written down in integral transforms. Some need to be approximated via simulation. We will present some properties of the model and introduce a consistent likelihood-based estimator. |
10:20 – 10:30 | Coffee break |
10:30 – 10:55 | Rasmus Waagepetersen: Asymptotic normality for conditionally centered space-time processes Abstract: This talk presents a central limit theorem for a sequence of random fields that are conditionally centered, meaning that they have zero mean given the past. We consider both increasing time, increasing space and a combination of increasing space and time for obtaining asymptotic normality of a normalized sum of the variables of the random fields. Weak dependence properties are required. Conveniently, in the case of increasing space, it suffices to assume weak spatial dependence for each random field separately when conditioning on the past. We give examples of applications to estimating functions for space-time autoregressive processes and space-time point processes. |
10:55 – 11:20 | Ron de Wit: Data-driven risk management: Present and future applications for fire services Abstract: Abstract: Veiligheidsregio Twente and Twente fire brigade are briefly introduced in general and as potential users of the research on fire prediction models. Four present cases in which the use of data plays an important role are introduced: cost-benefits of public fire alarms, chimney fire prediction, station dashboards and the so called ‘Daily safety situation’. In addition, two potential cases for future application of data and prediction-models are discussed. The presentation will be concluded by looking forward into steps that will further improve data driven risk management for Veiligheidsregio Twente and the Twente fire brigade. |
11:20 – 11:25 | Closing |
11:30 – 12:30 | Lunch |
12:30 – 14:00 | Defence Changqing Lu (Waaier 4, room open at 12:00) |
14:00 – 15:00 | Reception |