Set a custom stoping criteria for Simulated Annealing

Hi all,

I am using the "simulannealbnd" function to solve my optimization problem. I want to set a custom stoping criteria for my code which is the program will terminate if there is no improvement in ojective function after 1000 iterations. I asked ChatGPT and it provided me this code (see below). However it failed to run. Do you have any idea on this problem? Thanks in advance!

options = optimoptions('simulannealbnd','OutputFcn', @myOutputFunction)

function stop = myOutputFunction(optimValues, state)
    persistent bestValue counter maxIterationsWithoutImprovement;

    if isempty(bestValue)
        bestValue = Inf;  % Assume minimization
        counter = 0;
        maxIterationsWithoutImprovement = 1000;
    end

    if strcmp(state, 'iter')  % Check only during iterations
        if optimValues.bestfval < bestValue
            bestValue = optimValues.bestfval;
            counter = 0;
        else
            counter = counter + 1;
        end

        if counter >= maxIterationsWithoutImprovement
            disp(['No improvement in the objective function value after ', ...
                  num2str(maxIterationsWithoutImprovement), ' iterations. Stopping...']);
            stop = true;
        else
            stop = false;
        end
    else
        stop = false;
    end
end

The command window show errors like this:

Error using Inversion>myOutputFunction
Too many input arguments.

Error in globaloptim.simulannealbnd.saoutput (line 39)
    [stop ,optnew , changed ] = feval(OutputFcns{i},optold,optimValues, ...

Error in globaloptim.simulannealbnd.saengine/callOutputPlotFunctions (line 60)
                    globaloptim.simulannealbnd.saoutput(options,optimvalues,state,problem);

Error in globaloptim.simulannealbnd.saengine (line 17)
callOutputPlotFunctions('init');

Error in globaloptim.simulannealbnd.driver (line 28)
solverData = globaloptim.simulannealbnd.saengine(solverData,problem,options);

Error in simulannealbnd (line 197)
    globaloptim.simulannealbnd.driver(FUN, x0, [], [], [], [], lb, ub, options, defaultopt);

Error in Inversion (line 20)
[x_SA,fval_SA,exitFlag_SA,output_SA] = simulannealbnd(fun,x0,lb,ub,options)

Answer 1

[R]

To set a stopping criteria for the "simulannealbnd" function, you can use the "MaxStallIterations" option. By setting it to 1000, the iterations will terminate if there is no improvement in the objective function after 1000 iterations. Here's an example of how to set this option:

options = optimoptions('simulannealbnd','MaxStallIterations',1000);
options.FunctionTolerance

ans = 1.0000e-06

fun = @dejong5fcn;
x0 = [0 0];
x = simulannealbnd(fun,x0,[],[],options)

By modifying 'options.FunctionTolerance', you can set your custom stopping criteria. Refer to the following documentation for more information:

https://www.mathworks.com/help/gads/simulannealbnd.html#buy3g1g-2:~:text=integer%20%7C%20%7BInf%7D-,MaxStallIterations,-Number%20of%20iterations

[Ductho Le]

Hi @Raghvi,

Thank you for your comment. I have read about the 'MaxStallIterations' option but I couldn't understand what they mean in the description. Now, you make it clear. Thanks a lot!

Best,