Post doc position in numerical analysis at the Centre for Mathematical Sciences, Lund University
Description of the workplace
You will belong to a newly established group whose research is concerned with numerical methods connected to machine learning. We are financed by the Wallenberg AI, Autonomous Systems and Software Program (WASP).
In your closest surroundings at the Centre for Mathematical Sciences will be the research group for Numerical Analysis, who successfully do research with a focus on numerical methods for time integration. We are about 15 persons who do both pure theoretical analysis at a high level of abstraction as well as more practical software development.
Subject description
In this project, we will study and develop methods for so called neural differential equations, i.e. differential equations with many parameters. If these are to be used effectively, they require accurate, stable and fast methods for both forward- and backward-integration of the underlying differential equation, corresponding to evaluation and training of the network. You will help us develop new time stepping methods that are tailored for typical applications for both of these purposes, implement them, and analyze their convergence properties with a focus on how the trainable parameters affect these.