Speaker
Description
Plasma exists over a wide range of both temporal and spatial scales, going from the fusion reactors to the huge, and far away, galaxies. A better understanding of plasma physics is needed both for our fundamental understanding of the universe, but also to enable the development of new technologies.However, plasma modelling can be very challenging,due to its multi-scale nature.
Machine learning is enabling new ways to uncover models from data,but this comes at a great cost: the models may be very hard to understand. Using symbolic and sparse regression, it is now possible to extract interpretable models exclusively from data.
The goal is to apply sparse regression techniques to the plasma data, generated in a PIC(Particle-In-Cell) simulation, in order to uncover reduced models of nonlinear plasma dynamics, and compare them with known nonlinear plasma models, analysing the results and discussing their implications.
