ML SIGW0 - Revision history

Kresse at 14:36, 27 September 2024

2024-09-27T14:36:17Z

← Older revision		Revision as of 14:36, 27 September 2024
Line 7:		Line 7:
	The default value for {{TAG\|ML_MODE}}=''REFIT'' works reliably in most calculations, however, sometimes it is necessary to increase the regularization parameter to avoid instabilities during finite temperature molecular dynamics simulations.		The default value for {{TAG\|ML_MODE}}=''REFIT'' works reliably in most calculations, however, sometimes it is necessary to increase the regularization parameter to avoid instabilities during finite temperature molecular dynamics simulations.

	Suppose the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_MODE}}=''REFIT'' or {{TAG\|ML_IALGO_LINREG}}=4). The best is to control the regularization via this parameter and keep the noise parameter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1. Useful values for {{TAG\|ML_SIGW0}} are typically between 1E-1 and 1E-9. When the value is increased the regularization is increased. This often slightly increases the training set error but also reduces potential instabilities. Hence, if instabilities are observed during molecular-dynamics simulations, one should try to increase {{TAG\|ML_SIGW0}} to say 1E-2, then refit, and repeat the molecular-dynamics simulations.		Suppose the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_MODE}}=''REFIT'' or {{TAG\|ML_IALGO_LINREG}}=4). The best is to control the regularization via this parameter and keep the noise parameter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1. Useful values for {{TAG\|ML_SIGW0}} are typically between 1E-1 and 1E-9. When the value is increased the regularization is increased. This often slightly increases the training set error but also reduces potential instabilities. Hence, if instabilities are observed during molecular-dynamics simulations, one should try to increase {{TAG\|ML_SIGW0}} to say 1E-3, then refit, and repeat the molecular-dynamics simulations.

	More on the theory of this regularization parameter can be found in [[Machine learning force field: Theory#Regression\|this section]].		More on the theory of this regularization parameter can be found in [[Machine learning force field: Theory#Regression\|this section]].

Kresse at 14:34, 27 September 2024

2024-09-27T14:34:24Z

← Older revision		Revision as of 14:34, 27 September 2024
Line 7:		Line 7:
	The default value for {{TAG\|ML_MODE}}=''REFIT'' works reliably in most calculations, however, sometimes it is necessary to increase the regularization parameter to avoid instabilities during finite temperature molecular dynamics simulations.		The default value for {{TAG\|ML_MODE}}=''REFIT'' works reliably in most calculations, however, sometimes it is necessary to increase the regularization parameter to avoid instabilities during finite temperature molecular dynamics simulations.

	If the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_MODE}}=''REFIT'' or {{TAG\|ML_IALGO_LINREG}}=4)~~, the~~ best is to control the regularization via this parameter and keep the noise parameter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1. ~~Usefull~~ values for {{TAG\|ML_SIGW0}} are typically between 1E-2 and 1E-9. When the value is increased the regularization is increased. This often increases the training set error ~~slightly~~ but also reduces instabilities. If instabilities are observed during molecular dynamics simulations, one should ~~therefore~~ try to increase {{TAG\|ML_SIGW0}} to say 1E-2, refit, and repeat the molecular dynamics simulations.		Suppose the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_MODE}}=''REFIT'' or {{TAG\|ML_IALGO_LINREG}}=4). The best is to control the regularization via this parameter and keep the noise parameter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1. Useful values for {{TAG\|ML_SIGW0}} are typically between 1E-1 and 1E-9. When the value is increased the regularization is increased. This often slightly increases the training set error but also reduces potential instabilities. Hence, if instabilities are observed during molecular-dynamics simulations, one should try to increase {{TAG\|ML_SIGW0}} to say 1E-2, then refit, and repeat the molecular-dynamics simulations.

	More on the theory of this regularization parameter can be found in [[Machine learning force field: Theory#Regression\|this section]].		More on the theory of this regularization parameter can be found in [[Machine learning force field: Theory#Regression\|this section]].

Kresse: Describe useful parameter range. Point out instability issues.

2024-09-27T14:31:50Z

Describe useful parameter range. Point out instability issues.

← Older revision		Revision as of 14:31, 27 September 2024
Line 5:		Line 5:
	Description: This flag sets the precision parameter <math>s_{\mathrm{w}}</math> (see [[Machine learning force field: Theory#Bayesian linear regression\|here]] for definition) for the fitting in the machine learning force field method.		Description: This flag sets the precision parameter <math>s_{\mathrm{w}}</math> (see [[Machine learning force field: Theory#Bayesian linear regression\|here]] for definition) for the fitting in the machine learning force field method.
	----		----
	The default value for {{TAG\|ML_MODE}}=''REFIT'' works reliably in most calculations.		The default value for {{TAG\|ML_MODE}}=''REFIT'' works reliably in most calculations, however, sometimes it is necessary to increase the regularization parameter to avoid instabilities during finite temperature molecular dynamics simulations.

	~~However, if~~ the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_MODE}}=''REFIT'' or {{TAG\|ML_IALGO_LINREG}}=4), the best is to control the regularization via this parameter and keep the noise ~~paramter~~ <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1.		If the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_MODE}}=''REFIT'' or {{TAG\|ML_IALGO_LINREG}}=4), the best is to control the regularization via this parameter and keep the noise parameter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1. Usefull values for {{TAG\|ML_SIGW0}} are typically between 1E-2 and 1E-9. When the value is increased the regularization is increased. This often increases the training set error slightly but also reduces instabilities. If instabilities are observed during molecular dynamics simulations, one should therefore try to increase {{TAG\|ML_SIGW0}} to say 1E-2, refit, and repeat the molecular dynamics simulations.

	~~For~~ the theory of this regularization parameter ~~see~~ [[Machine learning force field: Theory#Regression\|this section]].		More on the theory of this regularization parameter can be found in [[Machine learning force field: Theory#Regression\|this section]].
	== Related tags and articles ==		== Related tags and articles ==
	{{TAG\|ML_LMLFF}}, {{TAG\|ML_MODE}}, {{TAG\|ML_IREG}}, {{TAG\|ML_SIGV0}}, {{TAG\|ML_IALGO_LINREG}}		{{TAG\|ML_LMLFF}}, {{TAG\|ML_MODE}}, {{TAG\|ML_IREG}}, {{TAG\|ML_SIGV0}}, {{TAG\|ML_IALGO_LINREG}}

Tbucko: small typo corrected

2024-08-07T17:13:30Z

small typo corrected

← Older revision		Revision as of 17:13, 7 August 2024
Line 5:		Line 5:
	Description: This flag sets the precision parameter <math>s_{\mathrm{w}}</math> (see [[Machine learning force field: Theory#Bayesian linear regression\|here]] for definition) for the fitting in the machine learning force field method.		Description: This flag sets the precision parameter <math>s_{\mathrm{w}}</math> (see [[Machine learning force field: Theory#Bayesian linear regression\|here]] for definition) for the fitting in the machine learning force field method.
	----		----
	The default value for {{TAG\|ML_MODE}}=''REFIT'' works ~~reliable~~ in most calculations.		The default value for {{TAG\|ML_MODE}}=''REFIT'' works reliably in most calculations.

	However, if the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_MODE}}=''REFIT'' or {{TAG\|ML_IALGO_LINREG}}=4), the best is to control the regularization via this parameter and keep the noise paramter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1.		However, if the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_MODE}}=''REFIT'' or {{TAG\|ML_IALGO_LINREG}}=4), the best is to control the regularization via this parameter and keep the noise paramter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1.

Karsai at 16:05, 3 July 2023

2023-07-03T16:05:26Z

← Older revision		Revision as of 16:05, 3 July 2023
Line 3:		Line 3:
	{{DEF\|ML_SIGW0\|1E-7\|for {{TAG\|ML_MODE}} {{=}} REFIT\|1.0\|else}}		{{DEF\|ML_SIGW0\|1E-7\|for {{TAG\|ML_MODE}} {{=}} REFIT\|1.0\|else}}

	Description: This flag sets the precision parameter <math>s_{\mathrm{w}}</math> for the fitting in the machine learning force field method.		Description: This flag sets the precision parameter <math>s_{\mathrm{w}}</math> (see [[Machine learning force field: Theory#Bayesian linear regression\|here]] for definition) for the fitting in the machine learning force field method.
	----		----
	The default value for {{TAG\|ML_MODE}}=''REFIT'' works reliable in most calculations.		The default value for {{TAG\|ML_MODE}}=''REFIT'' works reliable in most calculations.

Karsai at 15:55, 3 July 2023

2023-07-03T15:55:57Z

← Older revision		Revision as of 15:55, 3 July 2023
Line 5:		Line 5:
	Description: This flag sets the precision parameter <math>s_{\mathrm{w}}</math> for the fitting in the machine learning force field method.		Description: This flag sets the precision parameter <math>s_{\mathrm{w}}</math> for the fitting in the machine learning force field method.
	----		----
	The default value works ~~empirically very good~~ in most calculations		The default value for {{TAG\|ML_MODE}}=''REFIT'' works reliable in most calculations.

	However, if the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_MODE}}=''REFIT'' or {{TAG\|ML_IALGO_LINREG}}=4), the best is to control the regularization via this parameter and keep the noise paramter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1.		However, if the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_MODE}}=''REFIT'' or {{TAG\|ML_IALGO_LINREG}}=4), the best is to control the regularization via this parameter and keep the noise paramter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1.

Karsai at 15:55, 3 July 2023

2023-07-03T15:55:05Z

← Older revision		Revision as of 15:55, 3 July 2023
Line 7:		Line 7:
	The default value works empirically very good in most calculations		The default value works empirically very good in most calculations

	However, if the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_MODE}}='''REFIT''' or {{TAG\|ML_IALGO_LINREG}}=4), the best is to control the regularization via this parameter and keep the noise paramter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1.		However, if the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_MODE}}=''REFIT'' or {{TAG\|ML_IALGO_LINREG}}=4), the best is to control the regularization via this parameter and keep the noise paramter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1.

	For the theory of this regularization parameter see [[Machine learning force field: Theory#Regression\|this section]].		For the theory of this regularization parameter see [[Machine learning force field: Theory#Regression\|this section]].

Karsai at 15:54, 3 July 2023

2023-07-03T15:54:55Z

← Older revision		Revision as of 15:54, 3 July 2023
Line 7:		Line 7:
	The default value works empirically very good in most calculations		The default value works empirically very good in most calculations

	However, if the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_IALGO_LINREG}}=4), the best is to control the regularization via this parameter and keep the noise paramter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1.		However, if the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_MODE}}='''REFIT''' or {{TAG\|ML_IALGO_LINREG}}=4), the best is to control the regularization via this parameter and keep the noise paramter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1.

	For the theory of this regularization parameter see [[Machine learning force field: Theory#Regression\|this section]].		For the theory of this regularization parameter see [[Machine learning force field: Theory#Regression\|this section]].

Karsai at 15:54, 3 July 2023

2023-07-03T15:54:07Z

← Older revision		Revision as of 15:54, 3 July 2023
Line 5:		Line 5:
	Description: This flag sets the precision parameter <math>s_{\mathrm{w}}</math> for the fitting in the machine learning force field method.		Description: This flag sets the precision parameter <math>s_{\mathrm{w}}</math> for the fitting in the machine learning force field method.
	----		----
			The default value works empirically very good in most calculations

	If the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_IALGO_LINREG}}=4), the best is to control the regularization via this parameter and keep the noise paramter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1.		However, if the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_IALGO_LINREG}}=4), the best is to control the regularization via this parameter and keep the noise paramter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1.

	For the theory of this regularization parameter see [[Machine learning force field: Theory#Regression\|this section]].		For the theory of this regularization parameter see [[Machine learning force field: Theory#Regression\|this section]].

Karsai at 15:46, 3 July 2023

2023-07-03T15:46:10Z

← Older revision		Revision as of 15:46, 3 July 2023
Line 3:		Line 3:
	{{DEF\|ML_SIGW0\|1E-7\|for {{TAG\|ML_MODE}} {{=}} REFIT\|1.0\|else}}		{{DEF\|ML_SIGW0\|1E-7\|for {{TAG\|ML_MODE}} {{=}} REFIT\|1.0\|else}}

	Description: This flag sets the ~~initial reversed and squared~~ precision parameter <math>~~\frac{1}{\sigma_~~{\mathrm{w~~}}^{2~~}}</math> in the machine learning force field method.		Description: This flag sets the precision parameter <math>s_{\mathrm{w}}</math> for the fitting in the machine learning force field method.
	----		----

	~~For details about~~ the ~~optimization of~~ this ~~regularization~~ parameter see ~~[[Machine learning force field: Theory#Bayesian error estimation~~\|~~this section]]~~.		If the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_IALGO_LINREG}}=4), the best is to control the regularization via this parameter and keep the noise paramter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1.

			For the theory of this regularization parameter see [[Machine learning force field: Theory#Regression\|this section]].
	== Related tags and articles ==		== Related tags and articles ==
	{{TAG\|ML_LMLFF}}, {{TAG\|ML_MODE}}, {{TAG\|ML_IREG}}, {{TAG\|ML_SIGV0}}		{{TAG\|ML_LMLFF}}, {{TAG\|ML_MODE}}, {{TAG\|ML_IREG}}, {{TAG\|ML_SIGV0}}, {{TAG\|ML_IALGO_LINREG}}

	{{sc\|ML_SIGW0\|Examples\|Examples that use this tag}}		{{sc\|ML_SIGW0\|Examples\|Examples that use this tag}}

← Older revision		Revision as of 14:36, 27 September 2024
Line 7:		Line 7:
	The default value for {{TAG\|ML_MODE}}=''REFIT'' works reliably in most calculations, however, sometimes it is necessary to increase the regularization parameter to avoid instabilities during finite temperature molecular dynamics simulations.		The default value for {{TAG\|ML_MODE}}=''REFIT'' works reliably in most calculations, however, sometimes it is necessary to increase the regularization parameter to avoid instabilities during finite temperature molecular dynamics simulations.

	Suppose the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_MODE}}=''REFIT'' or {{TAG\|ML_IALGO_LINREG}}=4). The best is to control the regularization via this parameter and keep the noise parameter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1. Useful values for {{TAG\|ML_SIGW0}} are typically between 1E-1 and 1E-9. When the value is increased the regularization is increased. This often slightly increases the training set error but also reduces potential instabilities. Hence, if instabilities are observed during molecular-dynamics simulations, one should try to increase {{TAG\|ML_SIGW0}} to say 1E-2, then refit, and repeat the molecular-dynamics simulations.		Suppose the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_MODE}}=''REFIT'' or {{TAG\|ML_IALGO_LINREG}}=4). The best is to control the regularization via this parameter and keep the noise parameter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1. Useful values for {{TAG\|ML_SIGW0}} are typically between 1E-1 and 1E-9. When the value is increased the regularization is increased. This often slightly increases the training set error but also reduces potential instabilities. Hence, if instabilities are observed during molecular-dynamics simulations, one should try to increase {{TAG\|ML_SIGW0}} to say 1E-3, then refit, and repeat the molecular-dynamics simulations.

	More on the theory of this regularization parameter can be found in [[Machine learning force field: Theory#Regression\|this section]].		More on the theory of this regularization parameter can be found in [[Machine learning force field: Theory#Regression\|this section]].

← Older revision		Revision as of 14:34, 27 September 2024
Line 7:		Line 7:
	The default value for {{TAG\|ML_MODE}}=''REFIT'' works reliably in most calculations, however, sometimes it is necessary to increase the regularization parameter to avoid instabilities during finite temperature molecular dynamics simulations.		The default value for {{TAG\|ML_MODE}}=''REFIT'' works reliably in most calculations, however, sometimes it is necessary to increase the regularization parameter to avoid instabilities during finite temperature molecular dynamics simulations.

	If the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_MODE}}=''REFIT'' or {{TAG\|ML_IALGO_LINREG}}=4)~~, the~~ best is to control the regularization via this parameter and keep the noise parameter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1. ~~Usefull~~ values for {{TAG\|ML_SIGW0}} are typically between 1E-2 and 1E-9. When the value is increased the regularization is increased. This often increases the training set error ~~slightly~~ but also reduces instabilities. If instabilities are observed during molecular dynamics simulations, one should ~~therefore~~ try to increase {{TAG\|ML_SIGW0}} to say 1E-2, refit, and repeat the molecular dynamics simulations.		Suppose the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_MODE}}=''REFIT'' or {{TAG\|ML_IALGO_LINREG}}=4). The best is to control the regularization via this parameter and keep the noise parameter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1. Useful values for {{TAG\|ML_SIGW0}} are typically between 1E-1 and 1E-9. When the value is increased the regularization is increased. This often slightly increases the training set error but also reduces potential instabilities. Hence, if instabilities are observed during molecular-dynamics simulations, one should try to increase {{TAG\|ML_SIGW0}} to say 1E-2, then refit, and repeat the molecular-dynamics simulations.

	More on the theory of this regularization parameter can be found in [[Machine learning force field: Theory#Regression\|this section]].		More on the theory of this regularization parameter can be found in [[Machine learning force field: Theory#Regression\|this section]].

← Older revision		Revision as of 14:31, 27 September 2024
Line 5:		Line 5:
	Description: This flag sets the precision parameter <math>s_{\mathrm{w}}</math> (see [[Machine learning force field: Theory#Bayesian linear regression\|here]] for definition) for the fitting in the machine learning force field method.		Description: This flag sets the precision parameter <math>s_{\mathrm{w}}</math> (see [[Machine learning force field: Theory#Bayesian linear regression\|here]] for definition) for the fitting in the machine learning force field method.
	----		----
	The default value for {{TAG\|ML_MODE}}=''REFIT'' works reliably in most calculations.		The default value for {{TAG\|ML_MODE}}=''REFIT'' works reliably in most calculations, however, sometimes it is necessary to increase the regularization parameter to avoid instabilities during finite temperature molecular dynamics simulations.

	~~However, if~~ the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_MODE}}=''REFIT'' or {{TAG\|ML_IALGO_LINREG}}=4), the best is to control the regularization via this parameter and keep the noise ~~paramter~~ <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1.		If the regularization needs to be controlled manually, like e.g. in the fitting via singular value decomposition ({{TAG\|ML_MODE}}=''REFIT'' or {{TAG\|ML_IALGO_LINREG}}=4), the best is to control the regularization via this parameter and keep the noise parameter <math>s_{\mathrm{v}}</math> (see {{TAG\|ML_SIGV0}}) constant at 1. Usefull values for {{TAG\|ML_SIGW0}} are typically between 1E-2 and 1E-9. When the value is increased the regularization is increased. This often increases the training set error slightly but also reduces instabilities. If instabilities are observed during molecular dynamics simulations, one should therefore try to increase {{TAG\|ML_SIGW0}} to say 1E-2, refit, and repeat the molecular dynamics simulations.

	~~For~~ the theory of this regularization parameter ~~see~~ [[Machine learning force field: Theory#Regression\|this section]].		More on the theory of this regularization parameter can be found in [[Machine learning force field: Theory#Regression\|this section]].
	== Related tags and articles ==		== Related tags and articles ==
	{{TAG\|ML_LMLFF}}, {{TAG\|ML_MODE}}, {{TAG\|ML_IREG}}, {{TAG\|ML_SIGV0}}, {{TAG\|ML_IALGO_LINREG}}		{{TAG\|ML_LMLFF}}, {{TAG\|ML_MODE}}, {{TAG\|ML_IREG}}, {{TAG\|ML_SIGV0}}, {{TAG\|ML_IALGO_LINREG}}