Using relative tolerance with fmincon

Hi

As per the following post, the tolerances for fmincon are absolute. I would like to know if I can use relative tolerances, which is useful if the optimization parameters have different scales. If this is not possible, shall I scale the parameters manually? Or use the ScaleProblem option?

https://www.mathworks.com/matlabcentral/answers/93068-how-do-the-tolerances-work-in-the-optimization-toolbox

[Hesham]

The following document gives more information, but it actually contradicts with the previous one.

https://www.mathworks.com/help/optim/ug/tolerance-details.html?s_tid=answers_rc2-1_p4_MLT

Can you please update the document that stated that tolerances are absolute (it was posted by MATLAB staff)?

Answer 1

[Catalytic]

Can you please update the document that stated that tolerances are absolute (it was posted by MATLAB staff)?

No, we cannot. Only Matlab staff can update things posted by Matlab staff.