exit flags for lsqnonlin

Can anyone explain to me the difference between exitflags 1, 2 and 3 in lsqnonlin? I think my confusion here stems from the vague ways tolFun and tolX are described in the help files, but I'm not sure if there's more to my confusion.

Thanks,

David

[dpb]

For default of trust-region-reflective option it's in snls.m and the pertinent tests are--

if ((optnrm < tol1) && (posdef == 1) )
     ex = 1; EXITFLAG = 1;
     ...
 elseif (nrmsx < .9*delta) && (ratio > .25) && (diff < tol1*(1+abs(oval)))
     ex = 2; EXITFLAG = 3;
     ...
 elseif (iter > 1) && (nrmsx < tol2),
     ex = 3; EXITFLAG = 2;
     ...

I didn't dig thru the lower levels where some of these are defined...it certainly isn't clear from the doc's just what the difference is in any meaningful fashion, agreed...

Answer 1

[Matt J]

exitflag=1 means that the first order optimality measure was smaller than what the code thinks is significant.

exitflag=2 means that the change in x at the last iteration was smaller than what you consider significant, as expressed through your choice of TolX.

exitflag=3 means the change in the objective function components were smaller than what you consider significant, as expressed through your choice of TolFun.