Skip to content

Code repository for the paper "Multi-Fidelity Machine Learning for Molecular Excitation Energies" - Authors: Vinod, Vivin; Maity, Sayan; Zaspel, Peter; Kleinekathöfer, Ulrich

License

Notifications You must be signed in to change notification settings

SM4DA/MFML_for_ExcitationEnergies

Repository files navigation

Multi-Fidelity Machine Learning

(For Molecular Excitation Energy)

This repository contains scripts to reproduce the various results of the research paper (available at 10.1021/acs.jctc.3c00882).

The scripts are written in python. The files requirements.txt lists the versions of the libraries used. The rawdata is stored in the directories Data and Evaluation.

The scripts perform the following:

  • MFML_LearningCurves.py generates the outputs for all the the molecules and trajectories. These are used to generate the plots of Figure 3, Figure S5, and the scatter plots of Figure S7. The same outputs are used to generate Figure 4.
  • The script MFML_QZVP_LC.py generates the outputs used to generate Figure 5.
  • MFML_AntDFTB_analysis.py generates the output to generate Figure S6.
  • SingleDifferenceML_LearningCurves.py generates the two-level ML model as seen in Figure S8.

All plots can be generated via the jupyter notebook after suitable directory modifications while loading data.

About

Code repository for the paper "Multi-Fidelity Machine Learning for Molecular Excitation Energies" - Authors: Vinod, Vivin; Maity, Sayan; Zaspel, Peter; Kleinekathöfer, Ulrich

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published