MLatom@XACS is a powerful tool for training and using machine learning potentials. It supports a wide variety of representative potentials.

These potentials include:

·          Equivariant neural network MACE

·          Popular ANI with a good cost/performance ratio

·          Kernel methods such as KREG and sGDML which are fast and relatively good, particularly for small data

·          And many others

You can train and optimize the hyperparameters of these machine learning potentials using flexible options.

We also show how to benchmark the potentials for your application.

Once trained, you can use the potentials for your simulations such as dynamics or geometry optimization.

All of this you can do in a way you like the most, for example, in the Jupyter notebook or by submitting the input file.

We provide explanations, important tips, and sample notebooks and input files in the tutorial at https://xacs.xmu.edu.cn/docs/mlatom/tutorial_mlp.html