lsqnonlin different results version r2011b vs r2012a

I have implemented a program on Matlab r2012a. I have now to work with this code on the older version r2011b but the optimization algorithm lsqnonlin does not give me the same results.. How can I fix it?

[Matt J]

Possibly, you have a continuum of non-unique solutions. Are the resnorms the same in both versions, or are they different too? What is the magnitude of the difference you are seeing in the x and resnorms? What happens if you bring the initial point slightly closer to the desired solution?

Answer 1

[Shashank Prasanna]

Look under R2012a, there have been changes to the levenberg-marquardt algorithm. Here is the link to the release notes. You might have to to use a different algorithm:

http://www.mathworks.com/help/optim/release-notes.html