Lower and Upper bounds in LSQNONLIN

I am currently using the LSQNONLIN method in order to minimise an equation of n variables using least square.

For the variable I want to give the values of the lower and upper bounds as 1 and -1 respectively.

but during the optimisation in the toolbox, if i provide the starting point of the variable as a positive number between 0 and 1, then it will continue to optimize till the value of the variable is near 0, but does not then later take a negative value for the variable.

And if i provide a negative start point, thenthe value is not altered only.

What can i do in order to make it search for a solution in the entire range of -1 to 1.

Answer 1

[Steve Grikschat]

The start point is one of a few things you can change to influence the solver. There are other options (tolerances, etc) that may change the behavior as well.

However, it sounds like there may be multiple minima in your problem. Which of the different start points (positive or negative) gives the smallest residual?

You may opt to try many different start points to find the best solution.