Category:Machine-learned force fields - Revision history

Singraber: /* How to */

2026-03-31T22:15:55Z

How to

← Older revision		Revision as of 22:15, 31 March 2026
Line 17:		Line 17:
	*{{TAG\|Liquid Si - MLFF}}.		*{{TAG\|Liquid Si - MLFF}}.
	*[[Running machine-learned force fields in LAMMPS]]		*[[Running machine-learned force fields in LAMMPS]]
			*[[Running GRACE force fields in VASP]]

	=== Input ===		=== Input ===

Csheldon at 13:43, 24 March 2026

2026-03-24T13:43:03Z

← Older revision		Revision as of 13:43, 24 March 2026
Line 1:		Line 1:
	'''Machine-learned force fields''' used in combination with ab-initio [[:Category:Molecular dynamics\|molecular dynamics]] (MD) allow capturing the underlying physics from first principles and still reach long simulation times relatively cheaply. Generally, an ab-initio MD step is computationally expensive due to the [[:Category:Electronic minimization\|quantum-mechanical treatment of the electrons]], e.g., within density-functional theory (DFT). In fully classical MD calculations, '''force fields''' are used to evaluate the force acting on each atom instead of DFT. These interatomic potentials are traditionally based on experimental observation and empirical inclusion of known forces, such as the van-der-Waals force, electrostatic charges, etc. Therefore, the quality of the force field depends on how well interactions in the specific system are known.		'''Machine-learned force fields''' used in combination with ab-initio [[:Category:Molecular dynamics\|molecular dynamics]] (MD) allow capturing the underlying physics from first principles and still reach long simulation times relatively cheaply. Generally, an ab-initio MD step is computationally expensive due to the [[:Category:Electronic minimization\|quantum-mechanical treatment of the electrons]], e.g., within density-functional theory (DFT). In fully classical MD calculations, '''force fields''' are used to evaluate the force acting on each atom instead of DFT. These interatomic potentials are traditionally based on experimental observation and empirical inclusion of known forces, such as the van-der-Waals force, electrostatic charges, etc. Therefore, the quality of the force field depends on how well interactions in the specific system are known.

	VASP offers '''machine learning force fields on-the-fly''' to overcome these two issues. Namely, the high computational cost required for ab-initio MD and the empirical knowledge necessary to construct a force field the traditional way. To learn more about on-the-fly machine learning read about the [[Machine learning force field: Theory\|theory of on-the-fly machine learning force fields]] or about the [[Machine learning force field calculations: Basics\| setup of a basic calculation]]! Also, check out the description on [[Best practices for machine-learned force fields\|best practices on how to construct, test, and retrain force fields]], as well as the ~~[https~~:~~//www.vasp.at/tutorials/latest/md/~~part2/ basic tutorial that provides hands-on experience for silicon] and this tutorial: {{TAG\|Liquid Si - MLFF}}.		VASP offers '''machine learning force fields on-the-fly''' to overcome these two issues. Namely, the high computational cost required for ab-initio MD and the empirical knowledge necessary to construct a force field the traditional way. To learn more about on-the-fly machine learning read about the [[Machine learning force field: Theory\|theory of on-the-fly machine learning force fields]] or about the [[Machine learning force field calculations: Basics\| setup of a basic calculation]]! Also, check out the description on [[Best practices for machine-learned force fields\|best practices on how to construct, test, and retrain force fields]], as well as the {{Tutorial\|md:part2\|basic tutorial that provides hands-on experience for silicon}} and this tutorial: {{TAG\|Liquid Si - MLFF}}.

	== Theory ==		== Theory ==
Line 15:		Line 15:
	*[[Machine learning force field calculations: Basics\|Machine learning force field calculations: Basics]].		*[[Machine learning force field calculations: Basics\|Machine learning force field calculations: Basics]].
	*[[Best practices for machine-learned force fields]] include how to construct, test, and retrain force fields.		*[[Best practices for machine-learned force fields]] include how to construct, test, and retrain force fields.
	*[https://www.vasp.at/tutorials/latest/md/part2/ Basic tutorial that provides hands-on experience for silicon].
	*{{TAG\|Liquid Si - MLFF}}.		*{{TAG\|Liquid Si - MLFF}}.
	*[[Running machine-learned force fields in LAMMPS]]		*[[Running machine-learned force fields in LAMMPS]]
Line 41:		Line 40:
	, threshold and the total number of descriptors in the sparsification ({{TAG\|ML_EPS_LOW}}, {{TAG\|ML_RDES_SPARSDES}}). We recommend optimizing these in cross-validation by means of systematically varying the parameters and refitting the force field ({{TAG\|ML_MODE}}=refit). In principle, it is necessary to do that for all hyperparameters available in VASP to obtain the best MLFF model. In practice, usually the default values are a good guess. The default parameters of VASP were optimized on a selected set of bulk materials and on a molecular data base.		, threshold and the total number of descriptors in the sparsification ({{TAG\|ML_EPS_LOW}}, {{TAG\|ML_RDES_SPARSDES}}). We recommend optimizing these in cross-validation by means of systematically varying the parameters and refitting the force field ({{TAG\|ML_MODE}}=refit). In principle, it is necessary to do that for all hyperparameters available in VASP to obtain the best MLFF model. In practice, usually the default values are a good guess. The default parameters of VASP were optimized on a selected set of bulk materials and on a molecular data base.

	== Tutorials ==		== Additional resources ==
	*Tutorial for ~~[https~~:~~//vasp.at/tutorials/latest/md/~~part2/ training a machine learning force field using MD].		=== Tutorials ===
	*Tutorial for ~~[https~~:~~//vasp.at/tutorials/latest/mlff/~~part1/ error analysis and hyperparameter optimization].		*Tutorial for {{Tutorial\|md:part2\|training a machine learning force field using MD}}.
	*Lecture on the ~~[https~~:~~//youtu.be/f1VXKVlVhg4~~ basics of machine learning force fields].		*Tutorial for {{Tutorial\|mlff:part1\|error analysis and hyperparameter optimization}}.
	*Lecture on ~~[https~~:~~//youtu.be/HgGwgtkwh7g~~ machine learning force fields~~] including newer features.~~
	*Lecture on ~~[https~~:~~//youtu.be/bzzHpTBwxbA~~ chemical reactions using MLFFs].		=== Lectures ===
			*Lecture on the {{Video\|mlff:georg:2022\|basics of machine learning force fields}}
			*Lecture on {{Video\|mlff:ferenc:2025\|machine learning force fields}}
			*Lecture on {{Video\|chem_mlff:ferenc:2025\|chemical reactions using MLFFs}}



	[[Category:VASP\|Machine-learned force fields]][[Category:Ionic minimization\|Ionic minimization]][[Category:Molecular dynamics\|Molecular dynamics]][[Category:Forces\|Forces]]		[[Category:VASP\|Machine-learned force fields]][[Category:Ionic minimization\|Ionic minimization]][[Category:Molecular dynamics\|Molecular dynamics]][[Category:Forces\|Forces]]

Csheldon: /* Tutorials */

2025-10-20T13:06:48Z

Tutorials

← Older revision		Revision as of 13:06, 20 October 2025
Line 45:		Line 45:
	*Tutorial for [https://vasp.at/tutorials/latest/mlff/part1/ error analysis and hyperparameter optimization].		*Tutorial for [https://vasp.at/tutorials/latest/mlff/part1/ error analysis and hyperparameter optimization].
	*Lecture on the [https://youtu.be/f1VXKVlVhg4 basics of machine learning force fields].		*Lecture on the [https://youtu.be/f1VXKVlVhg4 basics of machine learning force fields].
	*~~Updated lecture~~ on [https://youtu.be/HgGwgtkwh7g machine learning force fields].		*Lecture on [https://youtu.be/HgGwgtkwh7g machine learning force fields] including newer features.
	*Lecture on [https://youtu.be/bzzHpTBwxbA chemical reactions using MLFFs].		*Lecture on [https://youtu.be/bzzHpTBwxbA chemical reactions using MLFFs].

Csheldon: /* Tutorials */

2025-10-20T13:06:14Z

Tutorials

← Older revision		Revision as of 13:06, 20 October 2025
Line 45:		Line 45:
	*Tutorial for [https://vasp.at/tutorials/latest/mlff/part1/ error analysis and hyperparameter optimization].		*Tutorial for [https://vasp.at/tutorials/latest/mlff/part1/ error analysis and hyperparameter optimization].
	*Lecture on the [https://youtu.be/f1VXKVlVhg4 basics of machine learning force fields].		*Lecture on the [https://youtu.be/f1VXKVlVhg4 basics of machine learning force fields].
			*Updated lecture on [https://youtu.be/HgGwgtkwh7g machine learning force fields].
	*Lecture on [https://youtu.be/bzzHpTBwxbA chemical reactions using MLFFs].		*Lecture on [https://youtu.be/bzzHpTBwxbA chemical reactions using MLFFs].



	[[Category:VASP\|Machine-learned force fields]][[Category:Ionic minimization\|Ionic minimization]][[Category:Molecular dynamics\|Molecular dynamics]][[Category:Forces\|Forces]]		[[Category:VASP\|Machine-learned force fields]][[Category:Ionic minimization\|Ionic minimization]][[Category:Molecular dynamics\|Molecular dynamics]][[Category:Forces\|Forces]]

Csheldon: /* Tutorials */

2025-09-22T10:15:29Z

Tutorials

← Older revision		Revision as of 10:15, 22 September 2025
Line 44:		Line 44:
	*Tutorial for [https://vasp.at/tutorials/latest/md/part2/ training a machine learning force field using MD].		*Tutorial for [https://vasp.at/tutorials/latest/md/part2/ training a machine learning force field using MD].
	*Tutorial for [https://vasp.at/tutorials/latest/mlff/part1/ error analysis and hyperparameter optimization].		*Tutorial for [https://vasp.at/tutorials/latest/mlff/part1/ error analysis and hyperparameter optimization].
	*Lecture ~~for~~ [https://youtu.be/f1VXKVlVhg4 basics of machine learning force fields].		*Lecture on the [https://youtu.be/f1VXKVlVhg4 basics of machine learning force fields].
			*Lecture on [https://youtu.be/bzzHpTBwxbA chemical reactions using MLFFs].


	[[Category:VASP\|Machine-learned force fields]][[Category:Ionic minimization\|Ionic minimization]][[Category:Molecular dynamics\|Molecular dynamics]][[Category:Forces\|Forces]]		[[Category:VASP\|Machine-learned force fields]][[Category:Ionic minimization\|Ionic minimization]][[Category:Molecular dynamics\|Molecular dynamics]][[Category:Forces\|Forces]]

Csheldon: /* Tutorials */

2025-09-22T10:06:29Z

Tutorials

← Older revision		Revision as of 10:06, 22 September 2025
Line 44:		Line 44:
	*Tutorial for [https://vasp.at/tutorials/latest/md/part2/ training a machine learning force field using MD].		*Tutorial for [https://vasp.at/tutorials/latest/md/part2/ training a machine learning force field using MD].
	*Tutorial for [https://vasp.at/tutorials/latest/mlff/part1/ error analysis and hyperparameter optimization].		*Tutorial for [https://vasp.at/tutorials/latest/mlff/part1/ error analysis and hyperparameter optimization].
			*Lecture for [https://youtu.be/f1VXKVlVhg4 basics of machine learning force fields].


	[[Category:VASP\|Machine-learned force fields]][[Category:Ionic minimization\|Ionic minimization]][[Category:Molecular dynamics\|Molecular dynamics]][[Category:Forces\|Forces]]		[[Category:VASP\|Machine-learned force fields]][[Category:Ionic minimization\|Ionic minimization]][[Category:Molecular dynamics\|Molecular dynamics]][[Category:Forces\|Forces]]

Csheldon: /* Tutorials */

2025-05-28T13:27:05Z

Tutorials

← Older revision		Revision as of 13:27, 28 May 2025
Line 42:		Line 42:

	== Tutorials ==		== Tutorials ==
	*Tutorial for [https://vasp.at/tutorials/latest/mlff/part1/ ~~MLFF calculations~~].		*Tutorial for [https://vasp.at/tutorials/latest/md/part2/ training a machine learning force field using MD].
			*Tutorial for [https://vasp.at/tutorials/latest/mlff/part1/ error analysis and hyperparameter optimization].


	[[Category:VASP\|Machine-learned force fields]][[Category:Ionic minimization\|Ionic minimization]][[Category:Molecular dynamics\|Molecular dynamics]][[Category:Forces\|Forces]]		[[Category:VASP\|Machine-learned force fields]][[Category:Ionic minimization\|Ionic minimization]][[Category:Molecular dynamics\|Molecular dynamics]][[Category:Forces\|Forces]]

Csheldon at 13:23, 28 May 2025

2025-05-28T13:23:27Z

← Older revision		Revision as of 13:23, 28 May 2025
Line 41:		Line 41:
	, threshold and the total number of descriptors in the sparsification ({{TAG\|ML_EPS_LOW}}, {{TAG\|ML_RDES_SPARSDES}}). We recommend optimizing these in cross-validation by means of systematically varying the parameters and refitting the force field ({{TAG\|ML_MODE}}=refit). In principle, it is necessary to do that for all hyperparameters available in VASP to obtain the best MLFF model. In practice, usually the default values are a good guess. The default parameters of VASP were optimized on a selected set of bulk materials and on a molecular data base.		, threshold and the total number of descriptors in the sparsification ({{TAG\|ML_EPS_LOW}}, {{TAG\|ML_RDES_SPARSDES}}). We recommend optimizing these in cross-validation by means of systematically varying the parameters and refitting the force field ({{TAG\|ML_MODE}}=refit). In principle, it is necessary to do that for all hyperparameters available in VASP to obtain the best MLFF model. In practice, usually the default values are a good guess. The default parameters of VASP were optimized on a selected set of bulk materials and on a molecular data base.

	~~----~~		== Tutorials ==
			*Tutorial for [https://vasp.at/tutorials/latest/mlff/part1/ MLFF calculations].

	[[Category:VASP\|Machine-learned force fields]][[Category:Ionic minimization\|Ionic minimization]][[Category:Molecular dynamics\|Molecular dynamics]][[Category:Forces\|Forces]]		[[Category:VASP\|Machine-learned force fields]][[Category:Ionic minimization\|Ionic minimization]][[Category:Molecular dynamics\|Molecular dynamics]][[Category:Forces\|Forces]]

Singraber: /* How to */

2025-01-24T08:37:40Z

How to

← Older revision		Revision as of 08:37, 24 January 2025
Line 17:		Line 17:
	*[https://www.vasp.at/tutorials/latest/md/part2/ Basic tutorial that provides hands-on experience for silicon].		*[https://www.vasp.at/tutorials/latest/md/part2/ Basic tutorial that provides hands-on experience for silicon].
	*{{TAG\|Liquid Si - MLFF}}.		*{{TAG\|Liquid Si - MLFF}}.
			*[[Running machine-learned force fields in LAMMPS]]

	=== Input ===		=== Input ===

Huebsch: /* Hyperparameters */

2023-10-04T09:31:11Z

Hyperparameters

← Older revision		Revision as of 09:31, 4 October 2023
Line 38:		Line 38:

	Hyperparameters are user-defined parameters of the MLFF model that will not be optimized during training of the MLFF. For instance, cutoff radii ({{TAG\|ML_RCUT1}}, {{TAG\|ML_RCUT2}}), the normalization and weighting of ab-initio training data ({{TAG\|ML_IWEIGHT}}, {{TAG\|ML_WTOTEN}}, {{TAG\|ML_WTIFOR}}, {{TAG\|ML_WTSIF}})		Hyperparameters are user-defined parameters of the MLFF model that will not be optimized during training of the MLFF. For instance, cutoff radii ({{TAG\|ML_RCUT1}}, {{TAG\|ML_RCUT2}}), the normalization and weighting of ab-initio training data ({{TAG\|ML_IWEIGHT}}, {{TAG\|ML_WTOTEN}}, {{TAG\|ML_WTIFOR}}, {{TAG\|ML_WTSIF}})
	, threshold and the total number of descriptors in the sparsification ({{TAG\|ML_EPS_LOW}}, {{TAG\|ML_RDES_SPARSDES}}). We recommend optimizing these in cross-validation by means of systematically varying the parameters and refitting the force field ({{TAG\|ML_MODE}}=refit).		, threshold and the total number of descriptors in the sparsification ({{TAG\|ML_EPS_LOW}}, {{TAG\|ML_RDES_SPARSDES}}). We recommend optimizing these in cross-validation by means of systematically varying the parameters and refitting the force field ({{TAG\|ML_MODE}}=refit). In principle, it is necessary to do that for all hyperparameters available in VASP to obtain the best MLFF model. In practice, usually the default values are a good guess. The default parameters of VASP were optimized on a selected set of bulk materials and on a molecular data base.

	----		----
	[[Category:VASP\|Machine-learned force fields]][[Category:Ionic minimization\|Ionic minimization]][[Category:Molecular dynamics\|Molecular dynamics]][[Category:Forces\|Forces]]		[[Category:VASP\|Machine-learned force fields]][[Category:Ionic minimization\|Ionic minimization]][[Category:Molecular dynamics\|Molecular dynamics]][[Category:Forces\|Forces]]