RuNNer has been written to a large extent by Jörg Behler. In late 2018 it has been released under the GPL3.0 license. If you are interested in RuNNer please contact Jörg Behler. Some other contributors to RuNNer are:

  • Jovan Jose Kochumannil Varghese: part of the pair symmetry functions and pair NNP approach.
  • Tobias Morawietz: Implementation of the screening of the electrostatic interactions at short distances and the Nguyen Widrow weights initialization.
  • Andreas Singraber: Some bugfixes and a performance optimization of the angular symmetry function type 3. Further Andi has written a C++ implementation of the HDNNP method in LAMMPs, which is fully compatible with RuNNer.
  • Michael Gastegger and Philipp Marquetand: Implementation of the element decoupled Kalman filter.
  • Tsz Wai Ko (Kenko): Implementation of the noise matrix for the Kalman filter and additional cutoff functions for compatibility with n2p2. Further Kenko has reconstructed the electrostatic part of RuNNer for constructing 3G-HDNNP based on fixed charge, environment dependent charges. In addition to this, Kenko has implemented 4G-HDNNP model in RuNNer.
  • Jonas A. Finkler: Implementation of the 4G-HDNNP and reconstruction of the electrostatic part of RuNNer with Kenko. Also contributed the random number generator 6 and the mergesort implementation. In addition, Jonas also implemented OPENMP parallization in RuNNer to optimize the speed of calculating symmetry functions.
  • Marco Eckhoff: Implementation of spin-dependent atom-centered symmetry functions to construct magnetic high-dimensional neural network potentials. Further, Marco developed the RuNNerActiveLearn tool.

The author is grateful to all his past and present group members, collaborators and users of RuNNer for discussions and feedback.