I am the main developer of the Atomic Energy Network (ænet), a free and open-source software package for the construction and usage of atomic interaction machine learning potentials (MLP) based on artificial neural networks (ANNs). In essence, ænet provides tools to train ANNs to the potential energy of atomic reference structures (Fig. 1). The package provides C and Fortran libraries that can be integrated in existing simulation software (e.g., LAMMPS and Tinker ) to actually use ANN potentials in atomistic simulations. Released in 2016, ænet has already found users and contributors in North America, Europe, and Asia. The user base is rapidly growing, and ænet’s reference paper is among the most downloaded articles of the journal Comput. Mater. Sci. .
N. Artrith*, and A. Urban, Comput. Mater. Sci., 114, 135 (2016) (Editor’s Choice)
See here a list of publications that made use of ænet.
