The winner of the NR prize this year is Johannes Voll Kolstø. He recently obtained his master degree in applied mathematics and physics with specialization in industrial mathematics.

The master thesis deals with support vector machines (SVM) on Riemannian manifolds. SVM are classically used for binary classification problems on Hilbert spaces. However, in many important applications, the data do not lie in a Hilbert space. Kolstø treats such data as belonging to a generalized surface, called a Riemannian manifold and develops a novel classification model based on SVM. The thesis spans many areas of industrial mathematics. It introduces the SVM in the classical setting, then combines a fair amount of differential geometry by optimization techniques to develop the main results of the thesis. Finally, it compares the proposed model with three other existing manifold SVM models on real world data.

The thesis is of very high quality and lies at the heart of the industrial mathematics area. The problem of how to use SVMs in a manifold setting is highly nontrivial and the methods developed are genuinely new. Kolstø shows strong ability to absorb difficult concepts from mathematical areas such as differential geometry. At the same time, he has been able to implement these new results in program code and done extensive testing and validation to ensure the applicability of the new methods to real world problems.

Kolstø is a worthy winner of this year’s NR prize.

The jury members were Brynjulf Owren and Sara Martino.