Phi-ML meets Engineering: On random grid neural processes for solving forward and inverse problems in parametric PDEs
Thursday 28 Sep 2023
Time: 13:00 - 14:00
Bio
Davor is a PhD student in Computational Geoscience at the University of Oslo in Norway. Prior to this he completed an MSc in Mathematical Modelling and Scientific Computing at the University of Oxford (2021) and a BSc in Geophysics at Imperial College London (2020). His main research interests are in using data, algorithms and mathematical models to address environmental challenges and develop sustainable solutions within science and engineering. He is keen to collaborate with others on such issues, particularly for low- and no-carbon energy and industrial solutions, but also for modelling in medicine and healthcare.
Research interests
During his time at the Turing, Davor is working on implementing machine learning models as an informed decision maker on how to deal with subgrid-scale processes within computational models. A key example of subgrid-scale processes are turbulent processes within (computational) fluid dynamics, but also include biological, chemical, geological, and various other processes throughout science and engineering. Two most common approaches to numerically dealing with these processes are through subgrid-scale modelling and the use of mesh adaptivity techniques. Machine learning allows us to use these fundamentally different approaches in combination, with the promise of a methodology that can outperform either when used in isolation. To achieve this, Davor’s research is in developing subgrid models that work on continually evolving grids and in training models to learn how discretization error relates to physics, the local solution fields, and local mesh shape and size.
