Re: Levenberg Marquardt algorithm

rahulfadnavis@xxxxxxxxx wrote: 
Hello,

I am wroking on a problem that requires me to use the levenberg
marquardt algorithm.

I am not a physics major and have never herd of any curve minimization
/ levemberg in my life before this day.

The probelem goes like this:

I have a set of points.

x - axis : Energy
y - axis : log of sensitivity

this gives me the initial curve and i make an initial guess.

i have a function f

Yt = ( ( sqr + 1) * ((1 + exp(alpha*(initialGuess - omega -
(energy))))) );
f = log(ys*(1/Yt));

This gives me another set of points based on my initial guess.

I have to minimize it.

But I have no clue of what is going...
I have the algorithm from this website:

http://www.ics.forth.gr/~lourakis/levmar/

If anoyone has ever seen this algo before or worked on a similar kind
of problem please help me out with this.

Desperately need help!

Thanks
Rahul


Hi Rahul,

For my university course in numerical computation we had to write a Matlab program to use the Levenberg-Marquardt algorithm to find the best fit to some data. It was written in Matlab and got the input data from an Excel file; simply a column of x data and a column of y data. You then make an initial guess as to the parameters of the model function and the program will find the best solution. If you want, I can send the program. It is tailored to fit to three gaussian peaks.
.

Relevant Pages

  • Re: Levenberg Marquardt algorithm
    ... > I am wroking on a problem that requires me to use the levenberg ... > marquardt algorithm. ... That algorithm is usually used to find the best curve fit to a set of data. ...
    (sci.math.num-analysis)
  • smaller errors after including correlations in the fit?
    ... If my fit (based on the Marquardt algorithm) ... where V is the covariance matrix and I add the statistical errors ... Prev by Date: ...
    (sci.stat.math)
  • HELP with errors using correlations
    ... If my fit (based on the Marquardt algorithm) ... where V is the covariance matrix and I add the statistical errors ... Prev by Date: ...
    (sci.stat.edu)
  • HELP with errors using correlations
    ... If my fit (based on the Marquardt algorithm) ... where V is the covariance matrix and I add the statistical errors ... Prev by Date: ...
    (sci.stat.edu)
Quantcast