Prior to the PhD defence of Robin Markwitz on the 26th of September 2025, of the thesis “Spatio-temporal point process models for interval-censored data”, there will be a symposium of both national and international speakers. Presentations will be given by Wouter Koolen (CWI, University of Twente), Frank van der Meulen (VU Amsterdam), Tobias Müller (University of Groningen), Claudia Redenbach (RPTU Kaiserslautern-Lindau) and Clara Stegehuis (University of Twente). The talks will cover various topics in statistics, probability and machine learning. A schedule can be found below.
The symposium is planned to run from 10:00 until 12:30 in Ravelijn 3334 with a 2-hour lunch break before the PhD defence begins at 14:30. There is no attendance fee, but registration is required. Please register here:
10:00 – 10:05 | Welcome and opening |
10:05 – 10:30 | Wouter Koolen - Adversarial learning of a time interval Abstract: We study the problem of learning a time interval in an online adversarial framework. In this talk we formalize the problem in game-theoretic finance, where it reads as follows: design a strategy for sequential trading in a single asset such that, if there is an interval where the asset price exhibits a large upcrossing, the strategy earns a most of that growth. We characterise precisely what can be achieved, and show how to do the learning efficiently. We make the connection to statistical testing, and in particular adjusters/calibrators for sequential preservation of evidence. |
10:30 – 10:55 | Frank van der Meulen - Stochastic Phylogenetic Models of Shape Abstract: Phylogenetic modeling of morphological shape in two or three dimensions is one of the most challenging statistical problems in evolutionary biology. As shape data are inherently correlated and non-linear, most naïve methods for phylogenetic analysis of morphological shape fail to capture the biological realities of evolving shapes. In this study we propose a novel framework for evolutionary analysis of morphological shape which facilitates stochastic character mapping on landmark shapes. The framework is based on recent advances in mathematical shape analysis and models the evolution of shape as a diffusion process that accounts for the evolutionary correlation between nearby landmarks. It uses a Metropolis-Hastings Markov Chain Monte Carlo sampling scheme for inferring ancestral shape and parameters of the model. We evaluate the new inference algorithm using simulations and show that the method leads to improved estimates of the shape at the root and well-calibrated credible sets of shapes at internal nodes. To illustrate the method, we also apply it to a previously published data set of butterfly wing images. |
10:55 – 11:10 | Coffee break |
11:10 – 11:35 | Tobias Müller - Poisson-Voronoi percolation, in high dimensions Abstract: In the talk, we will consider percolation on the Voronoi tessellation generated by a homogeneous Poisson point process on d-dimensional space. That is, with each point z of a Poisson point process (a certain, natural model of a random, countable subset of d-dimensional space) we associate its Voronoi cell C(z), consisting of all x that are closer to z than to any other point of the Poisson process. Each cell of the Voronoi tessellation is coloured black with probability p and white with probability 1-p (independently of all other cells). We say that percolation occurs if there exists an infinite, black connected cluster of cells, and the critical probability p_c is defined as the infimum of all values p for which the probability of percolation is positive. Famously, it was shown by Bollobás and Riordan in 2005 that p_c = 1/2 for Voronoi percolation in dimension d=2. In the talk I plan to discuss a result which states that, as d tends to infinity, we have p_c = (1+o(1))*(e/d)*2^{-d}. An adaptation of the proof also gives that for percolation on the Poisson-Gabriel graph we have p_c = (1+o(1))*2^{-d}. |
11:35 – 12:00 | Claudia Redenbach - Statistical tests for spatial point processes Abstract: Fitting a spatial point process model to an observed point pattern involves making various model decisions, which can be supported by statistical tests. These include goodness-of-fit tests for particular parametric models as well as tests for isotropy of the observed pattern, i.e., testing whether its distribution is invariant under rotations around the origin. Test statistics are often constructed from function-valued summary statistics such as Ripley's K-function, the nearest neighbour distance distribution function or their directional counterparts. In most cases, the exact distribution of the test statistics is not known. Therefore, asymptotic methods or Monte Carlo tests must be employed. In this talk, we will summarize the main steps and typical challenges of constructing tests. As a specific example, we will present a novel non-parametric and computationally cheap test for the hypothesis of isotropy for stationary point processes. Our test is based on resampling the Fry points which are the pairwise difference vectors of the observed point pattern. |
12:00 – 12:25 | Clara Stegehuis - Detecting geometry in scale-free networks Abstract: Geometric network models formalize the intuitive principle that similarity between vertices increases the likelihood of connection. As such, they successfully capture many structural features observed in real-world networks. However, if one observes only the network connections, the presence of geometry is not always evident. Traditional statistics such as triangle counts and clustering coefficients fail to detect geometry induced by hyperbolic spaces, or in networks with power-law degrees. To address these limitations, we introduce novel statistics based on weighted subgraph counts. These new tools are sensitive enough to detect geometry even in the weak geometry regime, where geometric effects start to vanish. |
12:25 – 12:30 | Closing |
12:30 – 14:30 | Lunch for committee members |
14:30 – 15:45 | PhD defence Robin Markwitz (Waaier Prof. dr. G. Berkhoff room, doors open at 14:00) |
15:45 – 16:45 | Reception @ coffee room, Zilverling 4rd floor |