Convergence of optimization routines

[Sigurður Ægir Jónsson]

Hi,

I have a very simple question, but for some reason I could not find an answer in the docs or in the forum (I must have missed it). When using the optimization routines, how do I check if they actually converged?

Here is the code snipped I am using:
double res = dlib::find_min_box_constrained(dlib::lbfgs_search_strategy(10),
dlib::objective_delta_stop_strategy(1.0e-8,1000),
obj_fun, dlib::derivative(obj_fun,1.0e-6), init_value, 0.0, 30.0);

[Davis]

Are you asking how you know if it terminated because of the 1e-8 condition
vs the 1000 iteration limit?

[Sigurður Ægir Jónsson]

Yes, that would be sufficient.

[Davis]

To do that you have to write your own version of the objective_delta_stop_strategy object that behaves in the way you want. If you look at the objective_delta_stop_strategy object it should be clear how to make your own.

[Sigurður Ægir Jónsson]

I will do that. Thank you very much.

[李岳容]

Hi, I would like to know how to find the source code of "objective_delta_stop_strategy" in order to make my own objective_delta_stop_strategy. I can only find the guide for using it in abstract.h. I must miss something. Thank you very much for your kind answer again.

[Davis]

Use git grep to find stuff like this