potfit wiki

open source force-matching

User Tools

Site Tools

uq

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision		Previous revision
Last revisionBoth sides next revision
uq [2018/11/27 16:12] – [Required parameters] slongbottomuq [2018/12/03 11:08] – [Hessian Bracketing Algorithm] slongbottom
Line 9: Line 9:
 <html><span style="color:red">This option is only available for analytic potentials.</span></html> <html><span style="color:red">This option is only available for analytic potentials.</span></html>
  
-This generates an ensemble of potentials whose spread can be used to quantify the uncertainties in the fitted parameters. Taking an uncorrelated subsample of the MCMC output forms a potential ensemble representing the uncertainties in each parameter by the ensemble spread and covariance. Propagating the uncertainty represented by the ensemble members through molecular dynamics, the resultant uncertainties in quantities of interest can be obtained. For an example of this see ((Longbottom, S., Brommer, P.: Uncertainty Quantification for Classical Effective Potentials. //archive link// )).+This generates an ensemble of potentials whose spread can be used to quantify the uncertainties in the fitted parameters. Taking an uncorrelated subsample of the MCMC output forms a potential ensemble representing the uncertainties in each parameter by the ensemble spread and covariance. Propagating the uncertainty represented by the ensemble members through molecular dynamics, the resultant uncertainties in quantities of interest can be obtained. For an example of this see ((Longbottom, S., Brommer, P.: Uncertainty Quantification for Classical Effective Potentials. //arxiv link// )).
  
 ===== The Ensemble Method ===== ===== The Ensemble Method =====
Line 20: Line 20:
 where $\lambda_j$ are the hessian eigenvalues, $V_{ij}$ the eigenvector components and $r_j$ is Gaussian noise. The R value, ''acc_rescaling'', is a tunable parameter for the MCMC step acceptance rate. where $\lambda_j$ are the hessian eigenvalues, $V_{ij}$ the eigenvector components and $r_j$ is Gaussian noise. The R value, ''acc_rescaling'', is a tunable parameter for the MCMC step acceptance rate.
  
-The MCMC algorithm samples potentials from the distribution at a temperature, $T_0$, set by the number of potential parameters and minimum cost value. In the majority of cases this temperature should be sufficient to generate a suitable ensemble. In the event that a reduced sampling temperature is required this can be scaled by a parameter $\alpha$ (''uq_temp''), such that $T=\alpha T_0$. +The MCMC algorithm samples potentials from the distribution at a temperature, $T_0$, set by the number of potential parameters and minimum cost value. In the majority of cases tis temperature should be sufficient to generate a suitable ensemble. In the event that a reduced sampling temperature is required this can be scaled by a parameter $\alpha$ (''uq_temp''), such that $T=\alpha T_0$. 
  
 +===== Hessian Bracketing Algorithm =====
 +
 +**Only use this if you know what you are doing! **
 +
 +If ''hess_pert = -1'' the parameter perturbations used in the finite difference calculation of the hessian are found individually. This algorithm can be used as a diagnostic tool to understand the curvature on the length scale of the sampling temperature. However care should be taken when analysing the information as many assumptions about the cost minimum are inherently made (e.g. that the landscape at the sampling temperature height is harmonic). Each parameter is perturbed to bracket the perturbation value yielding a the cost set by the sampling temperature - $C_T = C_0 + T = C_0 +\frac{2\alpha C_0}{N}$. When the bracketing interval is within 5% of $C_T$, a line is drawn between the two bounds and the gradient is used to choose the perturbation value estimated to give a cost of $C_T$.
 +
 +If the landscape at this scale is not harmonic, the eigenvalues of the hessian will be negative. In this case a reduced sampling temperature may be required and the user should think about improving the reference data being fit to, as well as the suitability and possible limitations of the potential model being used.
 ===== Parameters ===== ===== Parameters =====
  
 |<100% 33% 33% 33%>| |<100% 33% 33% 33%>|
 ^ parameter name | parameter type | default value | ^ parameter name | parameter type | default value |
-| short explanation |||+| short explanation. |||
  
 ====  Required parameters  ==== ====  Required parameters  ====
Line 53: Line 60:
  
 |<100% 33% 33% 33%>| |<100% 33% 33% 33%>|
-^ **hess_pert** | float | (none) | +^ **hess_pert** | float | 0.00001 | 
-| Percentage parameter perturbation in Hessian finite difference calculation. |||+| Percentage parameter perturbation in Hessian finite difference calculation. (If ''hess_pert = -1'' a bracketing algorithm is used to find individual parameter perturbation values, see explanation above - only use this is you know what you are doing!) |||
  
 |<100% 33% 33% 33%>| |<100% 33% 33% 33%>|
Line 60: Line 67:
 | Alternative MCMC step perturbation maximum value in max(''eig_max'', $\lambda_j$).||| | Alternative MCMC step perturbation maximum value in max(''eig_max'', $\lambda_j$).|||
  
 +|<100% 33% 33% 33%>|
 +^ **write_ensemble** | integer | 0 |
 +| Writes a potential file every ''write_ensemble'' members.|||