0

What is the expected format for the Grid Search input file?

When I tried to do a Grid Search with 4 parameters I got the message:

 

 *****  ERROR  *****
 * Evaluation of objective function currently only for a maximum of 3 active parameters!
 * Provide parameter sets to be evaluated on external file and use command: 
 * >>> GRID SEARCH from FILE: file_name
 *****  ERROR  *****

 

The Command Reference specifies that "An arbitrarily long list of parameter sets can be provided on file filename, each row consisting of one parameter set, whereby inactive and tied parameters are to be omitted."

But I cannot find anything else about the expected format of this input file for the Grid Search, and all my guessing attempts have been unsuccessful :(

 

Does this file requires only the number of intervals like the inverse file? Or the type and range of every paramater needs to be added in each row? Or it expects only the specific values to be evaluated for each parameter (in absolute or logarithm scale)? Does it recognize the inverse file keywords or the format is different?

 

I would appreciate if someone can explain or share an example.

 

Thanks in advance for the guidance!

3 replies

null
    • Finsterle GeoConsulting
    • Stefan_Finsterle
    • 3 wk ago
    • Reported - view

    Hi Héctor,

    Just provide a simple list of all parameter sets you want to be evaluated. For example, if you have four parameters specified in the > PARAMETER block, create a text file with four columns, each row holding the four values of that set, in the same order as specified in the > PARAMETER block. No need to indicate the type or range or number of sets (i.e., rows), which can be arbitrary, i.e., the parameters don't need to be aligned on a grid, but could be at any point in the parameter space of your choosing. If you want to have them on many points on a regular grid in the four-dimensional parameter space, just write a short program (or use EXCEL) to generate the list. 

    The list is read in free format.

    Yes, if you have the keyword >>>> LOGARITHM (or any other parameter transformation supported by iTOUGH2) in the > PARAMETER block, you have to specify the logarithm of that parameter.

    No need to specify any other keywords or commands (they are already given in the > PARAMETER block) - just the values in the same order.

    I'm aware that I'm not saying anything other than is already in the manual ("...each row consisting of one parameter set..."), and that you actually interpreted this simple scheme correctly ("...it expects only the specific values to be evaluated for each parameter..."). Still, I hope these additional comments are helpful. Good luck!

    Stefan

      • Hector_Ribera
      • 3 wk ago
      • Reported - view

       tried following the instructions but the simulation is always returning the following message:

       

      *****  WARNING  *****
       *  2 ambiguous keywords found near line  548:
       *      >>> GRID SEARCH from FILE : gs
       * Keyword No. 1 is interpreted!
       * --> 1 : GRID SEARCH
       *     2 : ROM
       *****  WARNING  *****

       

      You can find attached the grid search example input file, the inverse input and output files.

      • Finsterle GeoConsulting
      • Stefan_Finsterle
      • 3 wk ago
      • Reported - view

       Remove the word "from" from the command line, or add

      > Computation; >> STOP; >>> IGNORE WARNINGS (you should see that recommendation at the bottom of the iTOUGH2 output file).

      Stefan

Content aside

  • 3 wk agoLast active
  • 3Replies
  • 8Views
  • 2 Following