File for using KRR ML model to relax gold nanoclusters

Beskrivning

Using this file can construct KRR model, and then apply the KRR model to relax structures of Au25 clusters. requirement python code: numpy, dscribe(1.0.0 version), ase In data directory, Au25-0-Vacuum-fix-gold contains force fields files. 1, Constructing KRR and GPR models (our final model) In the "data" directory, "iniS.xyz", "ini_dftE", "ini_groE" are used for construnct the MBTR-KRR-FF-AL and MBTR-GPR-FF-AL model. To construct KRR and GPR model, it is need to use the bash script ("save_model.sh") to run "save_model.py" (sbatch save_model.sh), then the models are constructed and saved. 2, Mean absolute erro of energy To compute the MAE, we can use KRR model to predict energies. The function "calculate" in "gromacs_ml_calculator_gpr1.py" can be used to predict DFT energies. Then the prediction can compared with reference energy and force ("data/new_1000_test_dft" and "data/new_1000_test_gro"). 3, Using our saved model to relax structures In the "data" directory, we proveided 100 GROMACS relaxed structures and their DFT energies ("test_for_100relax.xyz" and "test_for_100relax"). To relax these 100 structures with our ML models, it is need to use the bash script ("runs.sh") to run "01-relax-test.py" (sbatch runs.sh).
Visa mer

Publiceringsår

2023

Typ av data

Upphovspersoner

Department of Applied Physics

Jarno Laakso Orcid -palvelun logo - Medarbetare

Patrick Rinke Orcid -palvelun logo - Medarbetare

Xi Chen Orcid -palvelun logo - Medarbetare

Lincan Fang - Upphovsperson

Zenodo - Utgivare

Projekt

Övriga uppgifter

Vetenskapsområden

Nanoteknologi

Språk

Öppen tillgång

Öppet

Licens

Creative Commons Attribution 4.0 International (CC BY 4.0)

Nyckelord

Ämnesord

Temporal täckning

undefined

Relaterade till denna forskningsdata