Phelixh/mixed-precision-sciml

Mixed precision scientific machine learning

Mixed Precision Scientific Machine Learning

The data and code for the paper J. Hayford, J. Goldman-Wetzler, E. Wang, & L. Lu. Speeding up and reducing memory usage for scientific machine learning via mixed precision. Computer Methods in Applied Mechanics and Engineering, 428, 117093, 2024.

Data

The dataset used for solving the problems associated with this paper.

• DeepONet examples
  • Advection equation
  • Linear Instability wave equation Refer to the original paper for this dataset: here
• PINN examples
  • Inverse problem of flow in a rectangular domain
  • Inverse problem of hemodynamics

Code

• DeepONet examples
  • LIW POD DeepONet
  • Physics-informed DeepONet
  • Advection equation
• PINN examples
  • Burgers equation
  • Kovasznay flow
  • Inverse Navier-Stokes
  • Hemodynamics
• Loss landscape and gradient analysis
  • Loss landscape
  • Training phases analysis
• Theoretical analysis of mixed precision accuracy
  • Theorem one validation

Cite this work

If you use this data or code for academic research, you are encouraged to cite the following paper:

@article{HAYFORD2024117093,
  author  = {Hayford, Joel and Goldman-Wetzler, Jacob and Wang, Eric and Lu, Lu},
  title   = {Speeding up and reducing memory usage for scientific machine learning via mixed precision},
  journal = {Computer Methods in Applied Mechanics and Engineering},
  volume  = {428},
  pages   = {117093},
  year    = {2024},
  doi     = {https://doi.org/10.1016/j.cma.2024.117093}
}

Questions

To get help on how to use the data or code, simply open an issue in the GitHub "Issues" section.