zur Startseite

Hauptseminar - (WS19/20)

Machine Learning in Numerical Simulation
Dozent Prof. Dr. rer. nat. habil. Miriam Mehl
M.Sc. Amin Totounferoush
M.Sc. Raphael Leiteritz
Umfang
Sprache
Studiengänge
Zielgruppe
Termine
Kurzbeschreibung

This seminar is aimed at students who are interested in using machine learning (specifically Artificial Neural Networks - ANN) to aid classical numerical simulation.

Deep neural networks have been proven to be capable of learning complex paradigms within data. This capability has been used in various fields to make predictions based on existing measured/gathered data. So far, numerical simulations have been carried out using the derived mathematical models. The purpose of using neural networks for numerical simulation is to learn the relationship between physical properties of different time steps just by using existing data instead of solving mathematical equations.

This introduces many challenges ranging from choosing the best-suited network architectures for specific problems to designing and implementing regularization strategies to increase the accuracy of predictions. There is a lot of research going on in this field which we want to cover in this seminar.

 

Conditions:

  • Fundamental understanding of numerical simulation

  • Active contribution to the discussions after the presentations

 

Procedure

  • Preliminary discussion with topic assignment

  • Submission of reports before the christmas break

  • Review process in the group and incorporation of the reviews

  • Final handover of the revised draft

  • Topic presentations