How to avoid unrecognized option names using optimget

I am using lsqcurvefit and receiving errors related to the function lsqncommon. I did not have these errors in the past, so this may be due to changes with more recent versions of Matlab. How can I edit the lines below to avoid these errors? Thank you.

Error 1. "Unrecognized option name 'InitDamping'. See OPTIMSET for possibilities."

Relevant lines of lsqncommon:

algorithm = optimget(options,'Algorithm',defaultopt,optimgetFlag);
initDamping = optimget(options,'InitDamping',defaultopt,optimgetFlag);
if ~iscell(algorithm)    
    initLMparam = initDamping; 
else
    initLMparam = algorithm{2}; % Initial Levenberg-Marquardt parameter
    algorithm = algorithm{1};   % Algorithm string
end

Error 2. "Unrecognized option name 'ProblemdefOptions'. See OPTIMSET for possibilities."

Relevant lines of lsqncommon:

ProblemdefOptions = optimget(options, 'ProblemdefOptions',defaultopt,optimgetFlag);
FromEqnSolve = false;
if ~isempty(ProblemdefOptions) && isfield(ProblemdefOptions, 'FromEqnSolve')
    FromEqnSolve = ProblemdefOptions.FromEqnSolve;
end

[Matt J]

You should not edit MathWorks-authored files. Show us the code that you wrote and are running, and let's discuss changes to that.

[Luca Martinetti]

Thank you for getting back to me. This is the structure of the code I am using that is related to lsqcurvefit. It worked with previous versions of Matlab in 2016-2019. I am seeing the errors above with the 2023 and 2024 versions. The errors above refer to specific lines of the function lsqncommon, not my own code. What do I have to do to avoid those errors? Thank you.

options = optimoptions('lsqcurvefit',                         ...
                       'Display', 'off',                      ...
                       'MaxFunctionEvaluations', MaxFunEvals, ...
                       'MaxIterations', MaxIter,              ...
                       'FunctionTolerance', TolFun,           ...
                       'StepTolerance', TolX,                 ...
                       'OptimalityTolerance', TolOpt,         ...
                       'Algorithm', 'trust-region-reflective');
[Parameters_Fit, RSS, residual, exitflag, output, ~, jacobian] =   ...
                   lsqcurvefit(function_handle, Parameters_Guess,  ...
                               X_data, Y_data,                     ...
                               Parameters_Lower, Parameters_Upper, ...
                               options);

[Walter Roberson]

I wonder whether those are options that are present only if you have Global Optimization Toolbox installed?

[Luca Martinetti]

It could be, I am not sure. Is there a way to find out?

[Matt J]

You need to provide a simplification of your code that we can run and reproduce the error. The code structure you've outlined looks fine, and works fine in the online Matlab engine, as demonstrated below,

[MaxFunEvals,MaxIter,TolFun,TolX,TolOpt]=deal(1e6,100,1e-6,1e-6,1e-6);

function_handle=@(p,xd) xd+p;
X_data=1:5;
Y_data=X_data+pi;
Parameters_Guess=3;
[Parameters_Lower, Parameters_Upper]=deal(0,10);

options = optimoptions('lsqcurvefit',                         ...
                       'Display', 'off',                      ...
                       'MaxFunctionEvaluations', MaxFunEvals, ...
                       'MaxIterations', MaxIter,              ...
                       'FunctionTolerance', TolFun,           ...
                       'StepTolerance', TolX,                 ...
                       'OptimalityTolerance', TolOpt,         ...
                       'Algorithm', 'trust-region-reflective');



[Parameters_Fit, RSS, residual, exitflag, output, ~, jacobian] =   ...
                   lsqcurvefit(function_handle, Parameters_Guess,  ...
                               X_data, Y_data,                     ...
                               Parameters_Lower, Parameters_Upper, ...
                               options)

Parameters_Fit = 3.1416

RSS = 1.8063e-25

residual = 1×5

1.0e-12 * -0.1901 -0.1901 -0.1901 -0.1901 -0.1901

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

exitflag = 1

output = struct with fields:
      firstorderopt: 6.5179e-12
         iterations: 3
          funcCount: 8
       cgiterations: 0
          algorithm: 'trust-region-reflective'
           stepsize: 4.3841e-07
            message: 'Local minimum found....'
       bestfeasible: []
    constrviolation: []

jacobian = 5x1 sparse double column vector (5 nonzeros)
   (1,1)       1.0000
   (2,1)       1.0000
   (3,1)       1.0000
   (4,1)       1.0000
   (5,1)       1.0000

[Walter Roberson]

Note that the user is using R2024a, but MATLAB Answers is using R2024b. The lsqncommon routine was rewritten between the two versions, and no longer has the problematic call.

[Matt J]

The lsqncommon routine was rewritten between the two versions, and no longer has the problematic call.

I can't find any documentation of that. I don't think we've even identified what the "problematic call" is, because we haven't seen the OP's actual code.

When I run the hypothetical code that I proposed in R2024a, it still runs fine.

[Luca Martinetti]

Thank you both. Although I am using R2024a, that particular code still uses an older (2016-2019) version of lsqcurvefit. This older version of lsqcurvefit gives me the same error when I try the simplified code above:

Error using optimget

Unrecognized option name 'InitDamping'. See OPTIMSET for possibilities.

Error in lsqncommon

initDamping = optimget(options,'InitDamping',defaultopt,optimgetFlag);

Error in lsqcurvefit

lsqncommon(funfcn,xCurrent,lb,ub,options,defaultopt,allDefaultOpts,caller,...

The current version of lsqcurvefit does not give me these errors, and the simplified code above runs just fine.

Something has changed in lsqcurvefit between 2016-2019 and now that is causing this different behavior.

Answer 1

[Matt J]

Although I am using R2024a, that particular code still uses an older (2016-2019) version of lsqcurvefit.

If so, problem solved. Use the version of lsqcurvefit that accompanies R2024a. It is obviously dangerous and unpredictable to mix Mathworks-authored files from different Matlab releases.

[Luca Martinetti]

Thanks