University of Münster honours the best doctoral graduates 2024
The Rectorate of the University of Münster honored the best doctoral graduates of 2024 on Friday (6 December). A total of 121 early career researchers received the highest distinction, "summa cum laude". To celebrate their achievements, the university's Rectorate hosted a reception at the "Schloss" in Münster. Among those honored from the Cluster of Excellence Mathematics Münster were Dr. Alessandro Codenotti, Dr. David Tobias Meyer, Dr. Philip Möller, Dr. Petr Naryshkin, and Dr. Adrian Riekert.
In addition, the Rectorate awarded the 2024 Dissertation Prize to 14 young researchers, including Dr. Adrian Riekert. Beyond scientific excellence, the dissertations must demonstrate a high degree of originality and make a significant contribution to current research. Rector Prof. Dr. Johannes Wessels, along with Vice-Rectors Prof. Dr. Maike Tietjens and Prof. Dr. Monika Stoll, presented the awards, each accompanied by €3,500. The prize money is intended to support further research by the awardees at the University of Münster or at another national or international institution.
Summary of Adrian Riekert’s dissertation
Topic: “Mathematical Analysis of Gradient Methods in the Training of Artificial Neural Networks”
Supervisor: Prof. Dr. Arnulf Jentzen
The dissertation focuses on algorithms for training artificial neural networks (ANNs). An ANN is a specific type of mathematical function composed of so-called artificial neurons. Each neuron receives various input signals and calculates an output value. This involves computing a weighted sum of the input values, followed by the application of an activation function. The weights of these sums are adjusted through a process called training to align the network's output with given data. While the methods used in practice are highly successful in many applications, a comprehensive theoretical justification for the convergence of training algorithms remains an open problem due to the complexity and high dimensionality of the functions involved. This dissertation proves, under certain assumptions, that the parameters of an ANN converge to a limit during training and that a trained ANN is highly likely to make accurate predictions for given data.
Links:
Mathematics Münster Graduate School
University of Münster Press Release on the "Summa cum laude" Event [de]
University of Münster Press Release on the Dissertation Prizes [de]