Capabilities include:

• State-of-the-art neural network models (HIPNN, ANI, etc.) for interatomic potentials for performing large-scale molecular dynamics in molecular and solid systems. https://github.com/lanl/hippynn
• Active Learning protocol for automatic generation of training dataset in the regions of large uncertainties.
• Training datasets (ANI, ANI-1x, ANI-1cc, etc.).
• PyTorch Semiempirical Quantum Mechanics (PySeQM) for training reduced quantum-mechanical models. https://github.com/lanl/PYSEQM

Contact: Sergei Tretiak

Research Highlights:
Teaching neural network to attach and detach electrons from molecules.
Zubatyuk, R.; Smith, J. S.; Nebgen, B. T.; Tretiak, S.; Isayev, O. Nature Communications 2021, 12 (1). doi.org/10.1038/s41467-021-24904-0

Extending machine learning beyond interatomic potentials for predicting molecular properties.
Fedik, N.; Zubatyuk, R.; Kulichenko, M.; Lubbers, N.; Smith, J. S.; Nebgen, B.; Messerly, R.; Li, Y. W.; Boldyrev, A. I.; Barros, K.; Isayev, O.; Tretiak, S. Nature Reviews Chemistry 2022, 6 (9), 653–672. doi.org/10.1038/s41570-022-00416-3

Deep Learning of Dynamically Responsive Chemical Hamiltonians with Semi-Empirical Quantum Mechanics.
Zhou, G.; Lubbers, N.; Barros, K.; Tretiak, S.; Nebgen, B. Proceedings of the National Academy of Sciences 2022, 119 (27). doi.org/10.1073/pnas.2120333119