Error "Too many output arguments" from fmincon

Hallo, everyone, I am using fmincon to solve an optimization problem. The main code is shown as below:

x = fmincon(@(x)e1_funcOptimal(x), x_0,[], [], [], [], -10, 10, @(x)e1_consOptimal(x), options);

where e1_funcOptimal(x) is the function provided the objective function for the optimization, and e1_consOptimal(x) is the constraint conditions which x need to satisfy.

The e1_funcOptimal(x) function is shown as below:

tol = 1.0E-13;
options = optimset( ...
    'Display',         'off', ...
    'GradObj',         'on', ...
    'GradConstr',      'on', ...
    'DerivativeCheck', 'off', ...
    'FinDiffType',     'central', ...
    'TolFun',          tol, ...
    'TolX',            tol, ...
    'TolCon',          tol, ...
    'algorithm',       'active-set', ...
    'MaxFunEvals',     inf, ...
    'MaxIter',         5000);

function obj = e1_funcOptimal(x)

obj = -x(1,10);
% or obj = -x(10)
% end

It means that I choose the negative value of one of the inputs as objective function.

The error messsage is that

Error using e1_funcOptimal
Too many output arguments.
Error in @(x)e1_funcOptimal(x)
Error in fmincon (line 561)
      [initVals.f,initVals.g] = feval(funfcn{3},X,varargin{:});
Caused by:
    Failure in initial objective function evaluation. FMINCON cannot continue.

How could I fix the problem. Thanks for reading.

[Adam Danz]

Could you show us the e1_funcOptimal() function? You showed us the e1_funcTariffOptimal() funciton instead.

[sxj]

hey, Adam, thanks for your notation. It's a typo and I have corrected it. That one is the function I choose to produce the objective function

[Walter Roberson]

Please show your options.

I predict that you configured in a way that requires a Jacobian to be output as the second output of your objective function.

[sxj]

hey, Walter, I have just upload the options. Do you mean I choose 'on' for the option "GradObj'' and 'GradConstr' and need to switch them to 'off'?

Answer 1

[Matt J]

Do you mean I choose 'on' for the option "GradObj'' and 'GradConstr' and need to switch them to 'off'?

Either switch them off or provide the gradients, e.g.,

function [obj,grad] = e1_funcOptimal(x)

     obj = -x(1,10);
     
     if nargout>1
       grad=zeros(size(x)); 
        grad(1,10)=-1;
     end
 
end

[sxj]

Thanks a lot, Matt. It works now. Have a good day