Parameter Tuning#

This subsection contains information regarding the tuning parameters with NOMAD. Enable tuning parameters requires the verbosity parameter in the subsection rpt parameters to be disabled by setting it as quiet otherwise it will interact with NOMAD since it needs the cost function value only. So far there are 3 types of cost functions implemented, one from Larachi et al. (1994) [1], the L1 function, and the L2 function. To tune parameters, the cost function compares the calculated counts with the Monte Carlo technique and the measured counts that are provided in the .experimental file. The three parameters dead time, activity and attenuation coefficient reactor seen in the detector parameter subsection of the .prm file are obtained using NOMAD. The second example Tuning Parameters with NOMAD explains how we can obtain the values of these parameters using NOMAD.

subsection parameter tuning
  set tuning                 = true
  set cost function type     = larachi
  set experimental data file = real_counts.experimental
end

tuning: Enable to tune parameters with NOMAD by showing the cost function in the terminal
Options: true or false (by default)
cost function type: Type of cost function to evaluate
Options: larachi (by default), l1 or l2
- Larachi cost function :
  
  \[f=\sum_{i=n}^{N}\left(\frac{C_i - M_i}{C_i + M_i}\right)^2\]
- L1 cost function :
  
  \[f=\frac{1}{N}\sum_{i=n}^{N}\left|C_i - M_i\right|\]
- L2 cost function :
  
  \[f=\frac{1}{N}\sum_{i=n}^{N}\left(C_i - M_i\right)^2\]
experimental data file: Filename of the text file with experimental/artificial counts
Options: Any text file with .experimental extension with the required header: experimental_counts

Parameter Tuning#

References#