/1d-pinn-reconstruction

This is the code for "Neural Network Reconstruction of Plasma Space-Time" by C.Bard and J.Dorelli (DOI: 10.3389/fspas.2021.732275). It is a Physics-Informed Transformer Neural Network which was used to reconstruct one-dimensional (M)HD shocktubes from partial samples. Includes source code, data, and jupyter notebooks for scientific reproduction

Primary LanguageJupyter NotebookOtherNOASSERTION

MHD_NN_reconstruction

This code is associated with the paper "Neural Network Reconstruction of Plasma Space-Time" by C.Bard and J. Dorelli (DOI: 10.3389/fspas.2021.732275)

Jupyter notebooks for reproducing paper plots are found in ./reproduce_plots ; the main programs are found in run_euler.py and run_mhd_recon.py.

The 'model' folders contain the best trained weights for each network. The 'spacetime' folders contain the raw HDF5 data from the baseline simulations.

Package Requirements: tensorflow, numpy, scipy, pylab, h5py

If you would like to cite this paper, the BibTex format is:

   @ARTICLE{2021FrASS...8..146B,
   author = {{Bard}, C. and {Dorelli}, J.~C.},       
    title = "{Neural Network Reconstruction of Plasma Space-Time}",        
  journal = {Frontiers in Astronomy and Space Sciences},      
 keywords = {Space Physics, reconstruction, Physics-informed neural network, MHD, computational methods},     
     year = 2021,         
    month = sep,        
   volume = {8},       
      eid = {146},          
    pages = {146},        
      doi = {10.3389/fspas.2021.732275},          
   adsurl = {https://ui.adsabs.harvard.edu/abs/2021FrASS...8..146B},       
  adsnote = {Provided by the SAO/NASA Astrophysics Data System}      
   }