Skip to content

Instructional implementation of Physics-Aware Training (PAT) with demonstrations on simulated experiments.

License

Notifications You must be signed in to change notification settings

KapilKhanal/Physics-Aware-Training

 
 

Repository files navigation

g5382

Physics-Aware Training (PAT) is a method to train real physical systems with backpropagation. It was introduced in Wright, Logan G. & Onodera, Tatsuhiro et al. (2022)1 to train Physical Neural Networks (PNNs) - neural networks whose building blocks are physical systems.

In this repository, we use examples based on simulated nonlinear coupled oscillators, to show how PNNs can be constructed and trained using PAT in PyTorch. Instead of a conventional python package, most of the code in this repository resides within self-contained Jupyter notebook examples. We have deliberately taken this approach, in the hopes that it will allow users to more easily understand and adapt this code for their own use. In our paper, we have taken essentially the same approach and demonstrated the methodology on real experiments.

Getting started

  • To learn about Physical Neural Networks, Physics-Aware Training, and the scope of this repository, have a look at the Introduction.
  • To see the examples that show how PNNs can be constructed and trained using PAT, see Examples.

How to cite this code

If you use Physics-Aware Training in your research, please consider citing the following paper:

Wright, L.G., Onodera, T., Stein, M.M. et al. Deep physical neural networks trained with backpropagation. Nature 601, 549–555 (2022). https://doi.org/10.1038/s41586-021-04223-6

License

The code in this repository is released under the following license:

Creative Commons Attribution 4.0 International

A copy of this license is given in this repository as license.txt.

About

Instructional implementation of Physics-Aware Training (PAT) with demonstrations on simulated experiments.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 99.9%
  • Python 0.1%