quantity#
Atomistic#
chemtrain provides machine learning routines for predicting atomistic
advanced atomistic properties. For example, the property_prediction
module transforms energy models to predict per-atom and per-molecule properties.
Macroscopic#
Ensemble theory connects the microscopic dynamics of a system to its macroscopic
properties via observables.
To enable perturbation-based training of these properties via DiffTRe,
chemtrain provides simple and advanced observables based on weighted
ensemble averages.
util.TargetBuilder provides a simple interface to initialize the
observables and compute functions simultaneously.