After resuming Bayesian optimization, the next sampling point is not the 'NextPoint' in the previous optimization results

Hello everyone.

I'm using bayesopt's 'resume' function to resume a previously ended Bayesian optimization(bayesopt). Here I use matlab2023a to calculate. The example adapts from the following link: https://ww2.mathworks.cn/help/stats/bayesianoptimization.resume.html

First, I performed 30 iterations and got the result which called 'results'. After resuming Bayesian optimization, I found that the sampling point to be evaluated immediately was not the variable 'NextPoint' in 'results', which means that the optimization did not continue according to the previous result. That is to say, performing 60 iterations at one time is different from performing 30 iterations first, and then using resume to continue 30 iterations. What is the reason for this phenomenon? Is there any good way to resume sampling from variable 'NextPoint' in last optimization result?

The codes I used are as follows:

x1 = optimizableVariable('x1',[-5,5]);
x2 = optimizableVariable('x2',[-5,5]);
vars = [x1,x2];
fun = @rosenbrocks;
rng default
results = bayesopt(fun,vars,'Verbose',0,'MaxObjectiveEvaluations',30,...
    'AcquisitionFunctionName','expected-improvement-plus','IsObjectiveDeterministic',true,'PlotFcn',[]);
results.NextPoint

ans = 1×2 table
       x1         x2   
    ________    _______

    -0.28356    0.11953

newresults = resume(results,'IsObjectiveDeterministic',true,'MaxObjectiveEvaluations',30,'PlotFcn',[]);
newresults.XTrace(31,:)

ans = 1×2 table
       x1         x2   
    ________    _______

    -0.28206    0.11514

function f = rosenbrocks(x)

f = 100*(x.x2 - x.x1^2)^2 + (1 - x.x1)^2;

end

As we can see from the above results,after using 'resume',the X point to evalate immediately is not same as the 'Nextpoint' in the 'results'.

[Alan Weiss]

I do not know the answer to your question. I have tried asking some other member of the development staff, but vacations abound right now. I hope that you will get an answer within a few weeks.

Alan Weiss

MATLAB mathematical toolbox documentation

[Nicholas]

Hi Alan

Thank you very much for your help! I am looking forward to the answer.

Best Wishes