Did anyone modify the 'Levenberg-Marquardt' VI ?

[Dmitry2002]

I'm using LabVIEW 6i. The subject VI parses model equation into a
certain table and all subsequent computation goes through this table.
The whole thing is quite time-consuming. Why not use a reference to a
VI with a formula node? It looks a little awkward and probably doesn't
fit the concept of LabVIEW, but this would take much less time.
I guess someone did get rid of the parsing and sped up the VI. So
could you please share the ideas on improving the VI or just share the
VI itself. Many people would appreciate this.
Thanks in advance, Dmitry.

[altenbach]

A few of us recently had a lenghty discussion about improvements to the currently available Lev-Mar routines.

David Thomson incorporated some of the ideas into the example available on his web site:

http://www.originalcode.com/occdl.html

Look For "Gaussian Peak Fit", see if this helps you get some ideas.

[JeanPierre]

I have posted a modification of the Lev-Mar target function here to call a function VI by reference. You are not forced to use a formula node for your computation. Implement it the way you want. The fit can also be speeded by putting the computation loop inside the function to avoid the overhead of calling a VI in each iteration.

[Dmitry2002]

Thanks, that helped me greatly.