Can bayesopt() be run without an objective function call?

Question

David Freiberg 2018-2-20

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/383808-can-bayesopt-be-run-without-an-objective-function-call

评论： xu supeng 2020-6-24

I have a list of initial X values and initial objectives, as well as variable ranges and parameters, but would like to just extract the next suggested point to sample with bayesObject('NextPoint'), because the evaluation is not a function I can easily put into MATLAB. I attempted to make a dummy function with the correct number of inputs and set 'MaxObjectiveEvaluations' to 0 to avoid any actual function calls:

   bayesObject = bayesopt(dummyFunc,var1,'InitialX',initialXList,'InitialObjective',initialObjList,'MaxObjectiveEvaluations',0);

but received the error

Error using bayesoptim.BayesoptOptions/checkAndFillStoppingCriteria (line X)
'MaxObjectiveEvaluations' must be a positive integer.

Can the 'NextPoint' property be called without ever calling the dummy function? Alternatively, is there a different function I should be using rather than bayesopt() for this?

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Don Mathis 2018-2-22

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/383808-can-bayesopt-be-run-without-an-objective-function-call#answer_306552

在 MATLAB Online 中打开

How about this?

The objective function returns 0 but is never called.
The initial points are counted as objective evaluations and so you can set MaxObjectiveEvaluations to 1 and no evaluations will actually occur.
Plotting and Verbose are turned off so it's fast and silent. (Takes about 0.1 seconds on my machine).

rng(0)
var1 = optimizableVariable('X', [-1,1]);
initialXList = table;
initialXList.X = rand(5,1)*2 - 1;
initialObjList = initialXList.X .^2;
dummyFunc = @(Tbl)0;
bayesObject = bayesopt(dummyFunc,var1,...
    'InitialX',initialXList,...
    'InitialObjective',initialObjList,...
    'MaxObjectiveEvaluations',1, ...
    'PlotFcn',{},...
    'Verbose',0);
bayesObject.NextPoint

2 个评论
显示无隐藏无

David Freiberg 2018-2-22

This worked perfectly - I didn't realize that 'InitialX' counted towards the number of function evaluations. Thank you very much.

xu supeng 2020-6-24

Could you help me to see my Question, I followed your method to write a program, but it seems the final result is related to the objective function value(for example if I set dummyFunc = @(Tbl)0 or dummyFunc = @(Tbl)100 the final results are different), but here you say the objective function is never called, here is my question,

https://www.mathworks.com/matlabcentral/answers/553393-how-to-use-bayesopt-function-to-predict-the-optimal-parameters-for-the-experiment

请先登录，再进行评论。