Overview
My research interests are at the intersection of deep learning and dynamical systems. This can mean one of two things. Either using deep learning methods for research related to dynamical systems theory. Or the opposite, using methods from dynamical systems theory for research related to deep learning.
I'm not currently affiliated with any research institution, so I'm working on a research project independently.
Scientific machine learning
Research topics like Neural Ordinary Differential Equations and Fourier Neural Operators makes it possible to learn dynamical systems from data.
This can be particularly useful for systems where its hard to explicitly model system dynamcs, in traditional machine learning fashion.
Physics-informed machine learning
Incorporating prior physical knowledge about a system derived from first principles can greatly simplify learning a system output, instead of relying solely on the dataset.
Physics-informed neural networks are one way to achieve this result, and also makes it possible to go beyong simply learning one particular system. For example by learning unknown or missing terms of an equation, or solving optimal control problems.
Peer reviewed papers
Work in progress
2025
Learning Ordinary Differential Equations with the Line Integral Loss Function
Albert Johannessen
2022
NeurIPS 2022 Workshop - The Symbiosis of Deep Learning and Differential Equations II
Master theses
Modeling Dynamical Systems with Physics Informed Neural Networks with Applications to PDE-Constrained Optimal Control Problems
Albert Johannessen
2024
NTNU Open
Motion Classification with Neural Ordinary Differential Equations
Albert Johannessen
2022
NTNU Open
I'm open for discussing concepts or collaborating on projects, so feel free to send me an email!
