How to display fmincon output outside the command windows ? (Deported output)

Hi,

I use Matlab fmincon opitmizer through matlab engine in an application developed in Qt.

I would like to know if there is a way to have the fmincon realtime output displayed outside the Matlab command Window ? Ideally deported in a Qt interface or at least on a maltab window that can be open by the run manager.

For all other displays it works already well but the fmincon output (and I assume that it is the same thing with other algorithm) is only displayed at the end of the whole optimization. In that way, user has no feedbak of the optimization process until its finished. It is disturbing when a processus can takes several hours !

Thanks a lot for your help !

Answer 1

[Matt J]

The way to gain access to status information during the execution of fmincon is by using the OutputFcn option,

https://www.mathworks.com/help/optim/ug/output-functions.html

https://www.mathworks.com/help/optim/ug/output-function.html#f11454

From within the Ouput Function, you can plot anything you want, anywhere you want, and use the status to trigger your own stopping criteria if you wish.

[Jérôme Vinçonneau]

Thanks a lot !

[Jérôme Vinçonneau]

I have an error message :

error using outfun too many output arguments :
Error in sqpInterface
Error in fmincon (line 823)
    [X,FVAL,EXITFLAG,OUTPUT,LAMBDA,GRAD,HESSIAN] = sqpInterface(funfcn,X,full(A),full(B),full(Aeq),full(Beq), ...

function outfun(optimValues, state)
switch state
    case 'init'
        hold on
    case 'iter'
        plot(optimValues.iteration,optimValues.fval)
        xlabel('iteration');
        ylabel('Objective function value');
        title('Fmincon info');
    case 'done'
        hold off
    otherwise
end

function [opt]=getOptimSolverOptions()
tolerance = 0.000000000001;
maxEval = 100000;
algo = 'sqp';

opt = optimoptions(@fmincon,'OutputFcn',@outfun,'Display','iter-detailed',... 
        'Algorithm',algo,'TolCon',tolerance,'CheckGradients',false,... 
        'Diagnostics', 'on', 'FiniteDifferenceType','central' ,... 
        'MaxFunEvals', maxEval, 'SpecifyObjectiveGradient', false); 
end

optimOptions = getOptimSolverOptions();

bMinOptim = true;
    [xMin,fvalMin,exitflagMin,outputMin,lambdaMin,gradMin,hessianMin] = fmincon(...
        @(x0)getCost(x0,bMinOptim),x0,A,b,Aeq,beq,lowBound,upBound,[],optimOptions);

[Matt J]

Jerome's comment moved here

The output and plot function have to be set by

opitmoptions('fmincon','OutputFcn', @outfun,'PlotFcn',@optimplotfval)

By reading the doc we can see that :

PlotFcn

You can use the PlotFcns option with the following MATLAB® optimization functions:

• fminbnd
• fminsearch
• fzero

The predefined plot functions for these optimization functions are:

• @optimplotx plots the current point
• @optimplotfval plots the function value
• @optimplotfunccount plots the function count (not available for fzero)

OutputFcn

You can use the OutputFcn option with the following MATLAB® optimization functions:

• fminbnd
• fminsearch
• fzero

However in the help of fmincon we have:

options = 
  fmincon options:

   Options used by current Algorithm ('interior-point'):
   (Other available algorithms: 'active-set', 'sqp', 'sqp-legacy', 'trust-region-reflective')

   Set properties:
     No options set.

   Default properties:
                    Algorithm: 'interior-point'
               CheckGradients: 0
          ConstraintTolerance: 1.0000e-06
                      Display: 'final'
     FiniteDifferenceStepSize: 'sqrt(eps)'
         FiniteDifferenceType: 'forward'
         HessianApproximation: 'bfgs'
                   HessianFcn: []
           HessianMultiplyFcn: []
                  HonorBounds: 1
       MaxFunctionEvaluations: 3000
                MaxIterations: 1000
               ObjectiveLimit: -1.0000e+20
          OptimalityTolerance: 1.0000e-06
                    OutputFcn: []
                      PlotFcn: []
                 ScaleProblem: 0
    SpecifyConstraintGradient: 0
     SpecifyObjectiveGradient: 0
                StepTolerance: 1.0000e-10
          SubproblemAlgorithm: 'factorization'
                     TypicalX: 'ones(numberOfVariables,1)'
                  UseParallel: 0
                  
                  

So it shoul work, no ?

[Matt J]

The format of the OutputFcn should be

function stop = outfun(x,optimValues,state)

but your version

function outfun(optimValues, state)

only has two input arguments and no output argument.

[Jérôme Vinçonneau]

yes but trying with the same synthax as the matlab example do not work too :

function stop = outfun(x,optimValues,state)
stop = false;
 
   switch state
       case 'init'
           hold on
       case 'iter'
           % Concatenate current point and objective function
           % value with history. x must be a row vector.
           history.fval = [history.fval; optimValues.fval];
           history.x = [history.x; x];
           % Concatenate current search direction with 
           % searchdir.
           searchdir = [searchdir;...
                        optimValues.searchdirection'];
           plot(x(1),x(2),'o');
           % Label points with iteration number.
           % Add .15 to x(1) to separate label from plotted 'o'
           text(x(1)+.15,x(2),num2str(optimValues.iteration));
       case 'done'
           hold off
       otherwise
   end
end

[Matt J]

With the same error message? I don't see why that would happen.

[Jérôme Vinçonneau]

finally it was a mistake in the wording and even if x is not used it is mandatory !!

Thanks a lot for your help