How to constrain lsqnonlin not to give me negative values???

Hi ! I am currently using the LSQNONLIN method in order to find the 4 variables of my equation that best fit on a know curve. I use lsqnonlin like this:

   X0=[0 4,5 0 T];
   options = optimset('Algorithm','levenberg-marquardt');   
   [fs,residual]=lsqnonlin(@OutInpDiff,X0,[],[],options);

My problem is that in the initial values X0; the fourth value (T) represents time delay. So when i run the program for different values sometimes for the best solution i get that the fourth parameter (T) have a negative value (T=-39.4521). That has no meaning since T is time delay.

So is there any way to constrain lsqnonlin to search for values which are greater than zero??

I tried

     [fs,residual]=lsqnonlin(@OutInpDiff,X0,0,[],options);

but i still get negative values... What can i do? Thanks

Answer 1

[Titus Edelhofer]

Hi,

you should define 0 as lower bound for all the variables, i.e.

[rs, res] = lsqnonline(@OutInpDiff, X0, zeros(1,4), [], options);

Or, if the other may be negative:

[rs, res] = lsqnonline(@OutInpDiff, X0, [-inf,-inf,-inf, 0], [], options);

Titus

[Walter Roberson]

Heh, I made exactly the same typo yesterday, lsqnonline in place of lsqnonlin

[Titus Edelhofer]

;-)