[bug #21953] Weird performance of grid search -- April 23, 2014 - 10:52

Update of bug #21953 (project relax):

                Priority:              5 - Normal => 7 - High               
                Severity:              3 - Normal => 4 - Important          
                  Status:                    None => Fixed                  
             Open/Closed:                    Open => Closed                 
        Operating System:        64-bit GNU/Linux => All systems            

    _______________________________________________________

Follow-up Comment #12:

The standard lower and upper values for kex was has been changed to between 1
and 10000.

Before, the standard lower and upper values for kex was set to 1 and 100000.

That means, that for a grid search with 10 GRID_INC, the kex valus would be
sought in intervals of
1, 10000, 20000 ...

For these guesses, the function value would be quite wrong, meaning that the
essential start point for kex would "normally" be 1.
This influenced the guess for the pA value, which in many cases would start at
0.5.

This bug was discovered for a CPMG data set, which previously was fitted to
the CR72 model and the start values of kex was far from the expected ~2000. 

    _______________________________________________________

Reply to this item at:

  <http://gna.org/bugs/?21953>

_______________________________________________
  Message sent via/by Gna!
  http://gna.org/