Use simulannealbnd without specifying the Upper andd lower bound

Is it possible to use the simulannealbnd without actually specifying the lowerbound and upper bound, yet adding ither options to the functions?

For ex for: [x,fval,exitflag,output] = simulannealbnd(fun,x0,lb,ub,options), I specified lb=[],ub=[]

I got error saying LB must be a double.

What could've been the problem?

Answer 1

[Walter Roberson]

use -inf and +inf for the bounds

Answer 2

[Sam Chak]

Hi @Subrat Bastola

There is no issue in this example. Perhaps, try clearing lb and ub variables from the Workspace.

fun = @(x) x.^4 - 16*x.^2 + 5*x;
fplot(fun), grid on, xlabel('x'), ylabel('f(x)')

x0  = 0;
% lb  = -5;
% ub  = 5;
options = optimoptions(@simulannealbnd, 'MaxIterations', 150)

options = 
  simulannealbnd options:

   Set properties:
             MaxIterations: 150

   Default properties:
             AcceptanceFcn: @acceptancesa
              AnnealingFcn: @annealingfast
                  DataType: 'double'
                   Display: 'final'
         FunctionTolerance: 1.0000e-06
                 HybridFcn: []
        InitialTemperature: 100
    MaxFunctionEvaluations: '3000*numberOfVariables'
        MaxStallIterations: '500*numberOfVariables'
                   MaxTime: Inf
            ObjectiveLimit: -Inf
                 OutputFcn: []
                   PlotFcn: []
          ReannealInterval: 100
            TemperatureFcn: @temperatureexp

% x   = simulannealbnd(fun, x0, lb, ub, options)
x   = simulannealbnd(fun, x0, [], [], options)

Maximum number of iterations exceeded: increase options.MaxIterations.

x = -2.9250