potfit wiki

open source force-matching

User Tools

Site Tools


output:ensemblefile

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
output:uq [2018/11/26 13:09] – [Format] slongbottomoutput:ensemblefile [2019/10/08 19:15] (current) – ↷ Page name changed from output:uq to output:ensemblefile slongbottom
Line 1: Line 1:
-====== Sloppyfile ======+====== Ensemblefile ======
  
 ---- ----
  
-The file output by the ensemble method (''sloppyfile''**.uq**) compilation contains the list of accepted MCMC potentials. +The ''ensemblefile.uq'' output by the uncertainty quantification ensemble method compilation contains the list of accepted MCMC potentials, see [[options:uq|uncertainty quantification]]
  
 =====  Format  ===== =====  Format  =====
  
-The file begins by outputting the calculated Hessian followed by rows listing its eigenvalues and eigenvectors in the format ''eigenvalue, eigenvector (e1,e2,...,eN)''. Then the accepted MCMC steps are listedwith the final three output columns for performance analysis purposes.+The file begins by outputting the calculated Hessian followed by a row listing its eigenvalues. This is followed by the corresponding matrix of eigenvectors, with the eigenvectors of each principle axes in each column, directly underneath it's related eigenvalue. Then the accepted MCMC steps are listedwith the final three output columns for performance analysis purposes.
  
 **id**\\ **id**\\
Line 13: Line 13:
  
 **param_1 param_2 ... param_N**\\ **param_1 param_2 ... param_N**\\
- the potential parameter set for each ensemble memeber+ the potential parameter set for each ensemble member
  
 **cost**\\ **cost**\\
Line 22: Line 22:
  
 **accepted**\\ **accepted**\\
- a flag of value ''1'' denotes that the step was accepted (users can have //potfit// output all unaccepted steps when compiling with the ''-debug'' option, with unsuccessful attempts referred to by a flag of ''0'' in this column)+ a flag of value ''1'' denotes that the step was accepted (users can have //potfit// output all unaccepted steps when compiling with the ''-debug'' option, see [[:options:main| compiling]], with unsuccessful attempts referred to by a flag of ''0'' in this column)
  
 **attempts**\\ **attempts**\\
- the total number of attempted MCMC moves, used to calculate the ''acc_prob''  (obviously when generating a potential ensemble users should let the Markov chain burn is to attain the correct acceptance probability before attempting to tune this using the R value, ''acceptance_rescaling'' parameter)+ the total number of attempted MCMC moves, used to calculate the ''acc_prob''  (obviously when generating a potential ensemble users should let the Markov chain burn is to attain the correct acceptance probability before attempting to tune this using the R value, ''acc_rescaling'' input parameter, see [[options:uq|input parameters]])
  
 **acc_prob**\\ **acc_prob**\\
Line 34: Line 34:
 An example output for a fictitious LJ potential fit is illustrated below:  An example output for a fictitious LJ potential fit is illustrated below: 
 <code> <code>
-#------------------------------------------------------ 
 # hessian: # hessian:
 # 804480.9 713749.4 # 804480.9 713749.4
 # 713749.4 1411542.1 # 713749.4 1411542.1
-#------------------------------------------------------ +
- +# eigenvalues: (l1, l2, ..., lN) 
-#------------------------------------------------------ +332402.6 1883620.4 
-eigenvalue, eigenvector (e1e2, ..., eN+# 
-332402.6  0.83406 -0.551658 +# eigenvectors; (v1v2, ..., vN
-1883620.4 0.55165  0.834069 +# 0.83406   0.55165 
-#------------------------------------------------------ +-0.551658 0.834069 
 +#
 # id param_1 param_2 cost weight accepted attempts acc_prob # id param_1 param_2 cost weight accepted attempts acc_prob
 0    0.6830  2.1120  62.5 10            10       0.10 0    0.6830  2.1120  62.5 10            10       0.10
output/ensemblefile.1543234180.txt.gz · Last modified: 2018/11/26 13:09 by slongbottom