Thread Subject: finding optimal starting points in lsqnonlin using RMSEARCH

Hi all,

I use the package RMSEARCH for finding starting points. It is a very good package and thank John for sharing this package with us. My question is as below.

I use lsqnonlin for finding optimal starting points for x where x is a 3*1 vector. RMSEARCH worked very well. But when I have a larger vector x, say, x is a 30*1 vector, then RMSEARCH didn't work. Is there any limit for the number of optimal parameters? How can I increase the limit?

I attached my code below. (x is 3*1 vector).

 opts = optimset('lsqnonlin');
 %opts.LargeScale = 'off';
 opts.Display = 'none';
 %opts.LevenbergMarquardt = 'off';
 opts.TolX = 1e-4;
 tic
[xfinal,ffinal,exitflag,xstart] = rmsearch(@(x) funtdlsq(x,shareGDPcal,vdata,difim4), ...
    'lsqnonlin',[0.2 0.744 0.67],[0 0.74 0.4],[1 0.7499 0.75],'initialsample',5,'options',opts)


Thank you very much in advance!

All my best,
Amber