What is MILADY?
MILADY - Machine Learning DYnamics is a package created by scientists for scientists, aiming to improve the accuracy and predictive power of atomistic simulations. The code uses machine learning algorithms and prioritises methods with a reasonable computational cost. MILADY is an open-source code distributed under the Academic Software Licence (ASL).
Main functionalities:
Development of machine learning force fields, suitable for molecular dynamics simulations.
Development of surrogate regression models to predict materials properties, including vibrational entropy, HOMO-LUMO energies, etc.
Representation of atomic structures in the feature space of atomic descriptors.
Analysis, sparsification, and optimization of datasets.
Advantages of MILADY:
MILADY is designed for High Performance Computing and is therefore well parallelized and has a good scalability. It uses MPI, PBLAS and ScaLapack.
MILADY includes a large choice of atomic descriptors and regression models, including linear, quadratic, polynomial chaos, and kernels.
MILADY is open-source package and contributions are very welcome.
Contributors
The coupling of MILADY and LAMMPS was developed by Thomas D. Swinburne (CINaM, Marseille).
Contact
Questions and suggestions can be sent to . We will be happy to answer!