Reference implementation of Molecular Orbital Based Machine Learning (MOB-ML) for electronic structure.
Any use of this code should cite the following paper:
"A Universal Density Matrix Functional from Molecular Orbital-Based Machine Learning: Transferability across Organic Molecules." L. Cheng, M. Welborn, A. S. Christensen, and T. F. Miller III, arXiv:1901.03309 (2019).
mobml_features.py generates MOB-ML features from the results of a HF calculation.
For documentation, see the docstrings within
mobml_features.py and especially the
feature_layout.md explains the layout of the feature vectors generated by
examples/ contains example inputs and outputs to
mobml_features.py in Matlab MATv5 format.
Each example contains inputs to the
a Fock matrix (
F), Coulomb matrix (
J), exchange matrix (
K), dipole matrices (
and the number of core (
nocc), and minimal basis orbitals (
Each example also contains reference outputs that should be generated by the
the MOB-ML features corresponding to diagonal (
and off-diagonal (
reference_offdiag_features) occupied LMO pairs (see Eqs. 5 and 6 of the above paper).
examples/water.mat: a single water molecule.