Constrained Gaussian processes
How can prior physical knowledge be incorporated into data-driven Gaussian process models?
Research
There are two main strategies to derive and deduce models — either using theory-based first principles or data-driven approaches. My overall research aim is to create new tools for using these two modeling strategies in conjunction. Most of the tracks below is financially supported by the Swedish Research Council (VR) via the project ‘‘Physics-informed machine learning’’ (registration number: 2021-04321)
For a complete list of publications, please refer to my Google Scholar.
Physics-informed neural networks
How can physics-informed neural networks be used in a probabilistic setting? 
Physics-informed generative models
We are developing physics-informed generative models tailored for scientific applications.
Magnetic localization
Encode physical prior knowledge in Gaussian processes