Link to my Google Scholar profile.
My postdoctoral research at TMU concerns an opinion dynamics model. We study this process on random geometric graphs to see how the geometry affects the ability of this process to reach a 'correct' majority or consensus.
An interactive visualization of the Louvain algorithm for community detection.
The focus of my PhD research was detecting communities in complex networks. Communities are groups of nodes that are better connected internally than externally. Such communities could, for example, correspond to groups of friends on social media or fields of study in citation networks. There are many heuristic algorithms for community detection, but there is limited understanding about which algorithm works well on what network. The most widely-used community detection method is modularity maximization. In this paper, we prove that this method is equivalent to minimizing an angular distance in a hyperspherical geometry. In another paper, we prove that several other community detection methods have the same hyperspherical structure.
For many machine learning tasks, there exist many measures to quantify the similarity between two outcomes. These measures are used to validate the performance of a machine learning algorithm by measuring the similarity between the outcome of the algorithm and the desired ('true') outcome. However, there are many of these validation measures, making it difficult to decide which one to use. This choice is especially important since many of these validation measures have inherent biases towards certain outcomes, which may lead to an unfair or suboptimal algorithm being chosen in practice.
For the machine learning tasks of clustering and classification, we mathematically analyzed the most popular validation measures, which led to recommendations about which measure to use in what setting. This resulted in two papers that were published at ICML2021 and NeurIPS2021, regarding the clustering and classification task respectively.
During the COVID-19 pandemic, we modeled the effect of mobility in the spread of the infections. This led to this publication in JRSI. An interview about this project can be found in the Cursor and a blog about this project (in Dutch) can be found on NEMO kennislink
My postdoctoral research concerns asynchronous majority dynamics on random graphs and is conducted under the supervision of Pawel Pralat.
I obtained my PhD under the supervision of Remco van der Hofstad and Nelly Litvak on the topic of community detection in random graphs.
For the KNAW (Dutch Royal Acadamy) and NEMO Kennislink, I wrote blogs as one of their Faces of Science. The blogs (in Dutch) are aimed at giving a glimpse into the life and work of young scientists. My profile can be found here.
For the Network Pages, a project by the Networks Center, I developed interactive visualizations that help explain network theory in a way that can be understood by non-mathematicians. An example of such a visualization is given below, taken from this blog about bottleneck detection algorithms.
As a student assistent, I assisted lecturers with various teaching tasks such as supervising student projects and grading assignments.
ViNotion is a specialist in the field of automated video content analysis that uses image recognition to analyze traffic. For ViNotion I used a game engine to generate synthetic datasets with ground truth. This data can be used to validate the performance of the products and to experiment with training on synthetic data. This video demonstrates the tool that I built.
For AthenaStudies, I gave exam trainings to help a group of first year Industrial Engineering students prepare for their mathematics resit.
With a specialization in Statistics, Probability and Operations Research. During this master, I did a 3-month internship at Yandex and MIPT in Moscow.
I also enjoy programming and software development. An example of one of my hobby projects is NetSweeper, a network-based variant of Minesweeper. The game is described in this blog for the NetworkPages and it can be played here.