Training Set

Selection of training set is an important step for developing a force field. The selection depends on the purpose. To parametrize valence and charge parameters, the training set should be molecular fragments that represents the atom types and interaction terms in target molecules. To optimize LJ parameters, the training set should be small molecules that have the atom types of interest.

Training set data are used for parametrization. For deriving valence and charge parameters, using the first-principle QM data of molecular fragments is undoubtedly the best option. For optimization of LJ parameters, high-quality data of liquid physical properties is essential.