Skip to content

Latest commit

 

History

History
23 lines (19 loc) · 1.18 KB

README.md

File metadata and controls

23 lines (19 loc) · 1.18 KB

Learning Physical Constraints with Neural Projections

Dependencies

All the dependencies are standard python packages:

  • Pytorch, Numpy, Matplotlib, PIL (used for visualization)

Code Structures:

  • _*.py: Some common classes/functions used for training and testing.
    • _constraint_net.py: A simple network used to represent the constraints to be learned
    • _iterative_proj.py: The iterative projection to solve the constrains
    • _training.py: Utils used for training
    • _dataloader.py: Load training data
    • _run_simulation.py: Run simulation use the learned constraint net and the projection operator
  • training_*.py: Train the model.
  • simulation_*.py: Use the trained model to generate simulations.
    • Simulation results will be written to \results folder.
    • In "evaluation" branch, simulation scripts are modified to run multiple simulation samples

Webpage:

https://y-sq.github.io/proj/neural_proj/