Re : Levenberg-Marquardt problem

[paganini]

hello

i want to fit my data with Levenberg-Marquardt function with a gaussian function model.

here is my vi ,, I ve got the error 20041 (singular matrix)

can you help me?

[altenbach]

Your model function only uses 2 parameters (look!), but you give initial estimates for 10 parameters. These sizes must match!

Also, you should always call the same model function by reference using the initial guesses so you can see if they are reasonable. Do you know reasonable initial guesses?

(Also note that you can delete the FOR loop in your model calculation without any change in result).

[altenbach]

OK, everything is wrong. 😞 Here is the right way to do it.

In addition to the above mentioned problem, your function misses one parameter, the ampltude. (you only use position and width). Attached is a fixed version. You need a function with three parameters and an array with three elements for the initial guess of the parametres.

Now the fit works just fine. The best fit is [Amplidude=~100, position=~1000, width=~3.2] 🙂

Let me know i you have any questions.

Message Edited by altenbach on 01-29-2007 01:29 PM

[paganini]

Thanks a lot for your quick response, you have made an excellent job.

Everything was wrong

Best regards

merci pour tout

Bonjour de la Guadeloupe!