Next: Appendices
Up: The Least-Squares Fitting Tool
Previous: The Working Dialogue
Use the Convergence Parameters dialogue to set the conditions
under which the fit will terminate. It is rarely useful to change
these settings from their default values and doing so is not
recommended. You may set the following values.
- Function Evaluation Limit
-
The maximum number of times that
may be evaluated, ignoring
evaluations for determining gradients and the curvature matrix.
- Iteration Limit
-
The maximum number of iterations to be performed. Note that
convergence will usually only take 10 to 20 iterations if the starting
model is reasonable.
- Relative Function Convergence Tolerance
-
The largest relative change in
that will be accepted as
showing that the process has converged. Other criteria must also be
met so that the program is sure that a true minimum has been reached.
Larger values (less than unity) are less stringent but raising this
value is rarely effective.
- X-Convergence Tolerance
-
The distance in scaled parameter space that is used to test whether
the parameters are close enough to their optimum values for the
process to have converged. The scaling factors vary dynamically
during the fit so this value can not be set from a set of accuracy
requirements. Large numbers are less stringent.
- False Convergence Tolerance
-
A parameter used in the false convergence test. Smaller numbers make
false convergence less likely but the default value is only 100 times
the machine precision, giving little scope for significant reductions.
Click on OK or Apply to make any changes take
effect. The dialogue will be dismissed if you click on OK
but will remain in place if you click on Apply.
Click on Cancel to dismiss the dialogue without changing the
convergence parameters.
Chris Flatters
Thu Mar 14 16:02:50 MST 1996