I keep getting "Error using / Arguments must be numeric, char, or logical. " in bayesopt.

Hi, I'm trying to use "bayesopt", but I keep getting this error.

What did I do wrong?

%% BayesOpt
fxc = @(x) cos(2*pi*(x/4))+exp(x/5);
xvar = optimizableVariable('x',[0,10],'Type','real');
res = bayesopt(fxc,xvar,'MaxObjectiveEvaluations',10,'PlotFcn',[],'AcquisitionFunctionName','lower-confidence-bound','ExplorationRatio',.1);
figure;
plot(res, @plotAcquisitionFunction);
figure;
plot(res,@plotObjectiveModel);
res=resume(res,'MaxObjectiveEvaluation',1);

Answer 1

[Walter Roberson]

doc bayesopt

fun accepts x, a 1-by-D table of variable values, and returns objective, a real scalar representing the objective function value fun(x).

Answer 2

[Jan]

Anonymous functions are compact, but less useful for debugging. Use a normal function instead:

fxc  = @yourFcn;
xvar = optimizableVariable('x',[0,10],'Type','real');
res  = bayesopt(fxc, xvar, 'MaxObjectiveEvaluations', 10, ...
    'PlotFcn', [], 'AcquisitionFunctionName', 'lower-confidence-bound', ...
    'ExplorationRatio', 0.1);
...
    
function y = yourFcn(x)
y = cos(2 * pi * x / 4) + exp(x / 5);
end

Run the code again. Now I'd expect it to fail inside yourFcn. Use the debugger to check the details:

dbstop if error

This let Matlab stop, when an error occurs. Run the code again, and when it stops, examine the class of x in the command window:

class(x)

What do you see?

[Walter Roberson]

I had to use dbstop if caught error

[Jan]

@DODO, @Walter Roberson: Yes, if the failing code is included in a TRY/CATCH block, "caught error" is the correct trigger to find the line causing the error.