This project is developed primarily by Andrew Peterson and Alireza Khorshidi in the Brown University School of Engineering. Specific credits:

  • Andrew Peterson: lead, PI, many modules
  • Alireza Khorshidi: many modules, Zernike descriptor
  • Zack Ulissi: tensorflow version of neural network
  • Muammar El Khatib: general contributions

We are also indebted to Nongnuch Artrith (MIT) and Pedro Felzenszwalb (Brown) for inspiration and technical discussion.


We would appreciate if you cite the below publication for any use of Amp or its methods:

Khorshidi & Peterson, “Amp: A modular approach to machine learning in atomistic simulations”, Computer Physics Communications 207:310-324, 2016. [doi:10.1016/j.cpc.2016.05.010]

If you use Amp for saddle-point searches or nudged elastic bands, please also cite:

Peterson, “Acceleration of saddle-point searches with machine learning”, Journal of Chemical Physics, 145:074106, 2016. [DOI:10.1063/1.4960708]

If you use Amp for uncertainty or with the bootstrap module, please also cite:

Peterson, Christensen, Khorshidi, “Addressing uncertainty in atomistic machine learning”, Physical Chemistry Chemical Physics, 19:10978, 2017. [DOI:10.1039/C7CP00375G]

If you use the initialization, feature and image selection, or nearsighted force training modules, we would appreaciate it if you can also cite:

Zeng, Chen and Peterson, “A nearsighted force-training approach to systematically generate training data for the machine learning of large atomic structures “. JCP 156, 064104 (2022). DOI:10.1063/5.0079314