MarieDeseyn / research-internship

Physics informed machine learning for AEM forward modeling (in the context of a 2 layer model)

Code for 4 machine learning techniques is available (in the corresponding folders):

• Simple auto-encoder
• Auto-encoder combined with Fourier transform
• CNN
• CNN combined with RNN
• basic code for CNN + auto-encoder combined with Fourier transform

For each method a general code for the technique (_gen) and an example code (without transformations, removal of large angles, optimization, etc. and therefore not very good models) in the AEM context (_ex) is present. The data used in the example codes is also available (Data_used_ex). For CNN, there is also a basic optimization code available (_opt), here each parameter is optimized individually, improvement might be found if the optimization is performed simultaneously for all parameters, however a low-RAM usage code should be written for this. The code for CNN + auto-encoder combined with Fourier transform is still very basic and needs improvement.

Data availability

The dataset used in this project is part of preliminary research. For the use of this data, please contact Wouter Deleersnyder.

Questions?

Contact us on GitHub!

• Wouter Deleersnyder
• Marie Deseyn