prony series conversion

[AXS]

HELLO

I am trying to make the algorithm of conversion ( experimental data to prony series param) using Matlab
I had some problem
Do you know if there is a function which allows to implement minimization algorithm
Do you have an script example using matlab to do it?

best regards

Pierre

[Jorgen]

If you use Matlab then it should be rather straightforward to create a program that does that conversion. The Matlab function that you are asking about is fminsearch (http://www.mathworks.com/access/helpdesk/help/techdoc/ref/index.html?/access/helpdesk/help/techdoc/ref/fminsearch.html).

I don't know if I have a polished script that I can share, I'll take a look...

Jorgen

[AXS]

Hello

As you said, i tried to use the fminsearch function in matlab
But it don't converge. I don t know if i have to sum the two error terms or use an other method:

the inputs are: xdata (=w); ydata1 (=G') ; ydata2 (=G")

function [estimates, model] = fitcurve(xdata, ydata2,ydata1)

% Call fminsearch with a starting point.

start_point = [69800,1,1,0001,000.1,000.1,000.1,000.1,00.1,0.1,0. 1,0.1,0.1,0.1,0.1,0.1,0.1,0.1,0.1,1,1,1,1,1,1,1,1, 1,1,0.1,0.1,1,1,1,1,1,1,1,1,1,1

model = @expfun;

options=optimset('MaxFunEvals',1E+25,'MaxIter',1E+ 25);

estimates = fminsearch(model, start_point,options);


function [sse, FittedCurve1] = expfun(params)

G0= params(1);
A1 = params(2);
lambda1 = params(3);
A2 = params(4);
lambda2 = params(5);

% Calculation of G"

numpart1= xdata.*lambda1.^2;
numpart2= xdata.*lambda2.^2;

G'' = G0*(( numpart1*A1 ./(lambda1.^2 * xdata.^2+1)+ numpart2*A2./(lambda2.^2 * xdata.^2+1)


% Calculation of G'

denumpart1=xdata.^2*lambda1.^2;
denumpart2=xdata.^2*lambda2.^2;


G' =G*((1-(A1+A2))+(denumpart1*A1 ./(lambda1.^2 * xdata.^2+1)+denumpart2*A2 ./(lambda2.^2 * xdata.^2+1));



FittedCurve1 = G";
FittedCurve2 = G';

ErrorVector1 = abs(FittedCurve1 - ydata2)+abs(FittedCurve2 -ydata1);;


sse = sum(ErrorVector1 .^ 2)
end
end

thanks for your help

best regards

Pierre

[Jorgen]

Here are a few comments:

- Your function at first glance looks OK. I would recommend, however, that you add comments to your code in order for other people to better understand what your approach!

- What is the start_point vector? Why it is so long??

- Do you really want Matlab to take 1e25 interations?? Why?

- I don't quite follow your calculation of G' and G", can you explain more?

- In general, lack of convergence is typically due to either a bad starting guess for the parameters, or an objective function that does not have a minimum.

Jorgen