What does this lsqnonlin error mean?

Hi

I'm attempting a nonlinear parameter fit using lsqnonlin and a Simulink model. I'm following a similarly method to: http://www.mathworks.com/help/optim/ug/lsqnonlin-with-a-simulink-model.html

After the function starts, I get this weird error below. What does it mean?

                                        First-Order                    Norm of 
   if true
    % Iteration  Func-count    Residual       optimality      Lambda           step
     0           3         231.377            51.4         0.01
     1           6         179.496            51.1        0.001        6.82145
Subscripted assignment dimension mismatch.

 Error in C:\MATLAB\R2012a Student\toolbox\shared\optimlib\finitedifferences.p>finitedifferences (line 159)

 Error in levenbergMarquardt (line 235)
            [JAC,~,~,numEvals,evalOK] = finitedifferences(XOUT,funfcn{3},[],[],[],costFun, ...

 Error in lsqncommon (line 181)
    [xC,FVAL,JACOB,EXITFLAG,OUTPUT,msgData] = ...

 Error in lsqnonlin (line 235)
[xCurrent,Resnorm,FVAL,EXITFLAG,OUTPUT,LAMBDA,JACOB] = ...

 Error in runSpringDamplsq (line 34)
param = lsqnonlin(@springDamplsq, param0, [], [], options);
  end

Answer 1

[Steve Grikschat]

This error is occurring because your function is returning arrays of varying sizes. In this case, it is being caught during finite-difference estimation of the function's Jacobian.

Check your function that you pass to lsqnonlin and make sure that it returns an array with a consistent number of elements.

[Amir Patel]

100% correct. I was stopping the simulation when end conditions were reached. This resulted in varying size arrays.

Thank you for the help!