Image of Martin Ritzert

About Me

I am a Junior Group Leader at ScaDS.AI at Leipzig University, the data science cluster of Leipzig University and TU Dresden. My research revolves around machine learning of graph data where I am interested in how to generate graphs (especially combined with other modalities such as 3D information and fine details), how to apply graph learning methods to traditional relational databases, and how to train message-passing networks (the main paradigm for graph learning) to approximate traditional algorithms for combinatorial problems such as the 3-coloring problem.

There is an open PhD position in my group on generative models for graphs. Please submit your application here: PhD in Graph Machine Learning.
If you have any questions regarding the position, just drop me an email.

Martin Ritzert

martin.ritzert@uni-leipzig.de

Graph Machine Learning Group
ScaDS.AI at Leipzig University
Leipzig, Germany

Google Scholar
LinkedIn

Teaching

I am currently looking for motivated students for bachelor and master thesis projects in the area of graph learning. Concrete thesis topics include the following:

  • Similarity Measures for botanical trees capturing 3D information and structure
  • Graph Transformers for neurons and botanical trees
  • Graph Drawing using Graph Machine Learning
  • GNNs with registers (generalizing virtual nodes)
If you are interested in one of these topics or have your own idea, please do not hesitate to contact me. Since ScaDS.AI is located in the Sparkassen-building you will need an appointment to enter the building, so please contact me via email to arrange a meeting.

Research

I am generally interested in machine learning and in particular machine learning on graph data. Graphs are a very general data structure and can be used to model many different types of data, such as social networks, molecules, and 3D scans of neurons or botanical trees. Some concrete topics are:

  • Graph Generation
  • Machine learning on relational databases
  • Learned approximation (message-passing) algorithms for NP-hard problems (e.g. 3-coloring)
  • Benchmarks for graph machine learning

I am further interested in complexity, learning theory, graph theory, and lately clustering, especially hierarchical clustering.

Research Background

Before joining Leipzig University as a Junior Group Leader, I was a Postdoc at Georg-August University Göttingen and Aarhus University. In Göttingen, I worked with Alexander Ecker on applying graph learning methods to 3D scans of neurons. The key question, we were trying to answer is how much the 3D morphology of a (mouse brain) neuron tells us about its function.

At Aarhus University, I worked with Kasper Green Larsen on the generalization performance of boosting methods. There, we showed that the popular AdaBoost method is not an optimal weak to strong learner, while a slight modification (using AdaBoost inside of AdaBoost as a weak learner) is optimal in terms of number of samples needed for learning.

During my PhD in Aachen under the supervision of Martin Grohe, I mainly worked on the algorithmic complexity of learning of logical formulas over graphs. Throughout my academic career, I have been working with practical graph learning in various smaller projects.

Selected Publications

For a full list of my papers, please visit my google scholar page.

Peer-reviewed (selection)

Short CV

  • since 2026 Junior Group Leader, ScaDS.AI, Leipzig University, Germany
  • 2022-2025 PostDoc, Computer Science, Georg-August University Göttingen, Germany
  • 2021-2022 PostDoc, Computer Science, Aarhus University, Denmark
  • 2016-2021 PhD, Informatik, RWTH Aachen University, Germany
  • 2014-2016 Master of Science, Informatik, RWTH Aachen University, Germany
  • 2011-2014 Bachelor of Science, Informatik, RWTH Aachen University, Germany
Plain Academic