Which algorithm is used in AnyOptStudy for optimization?
When I’ve worked with AnyOptStudy, I’ve cathed, that speed of optimization is very much depend from started approximation. What is the way to better choice of the started approximation vector?
Hi Pavel,
It is a feasible directions algorithm. The reason for this choice is that there is no simple way of doing efficient sensitivity analysis on the kind of analysis we perform in AnyBody. A sensitivity analysis wrt. one variable is basically as numerically expensive as a normal analysis.
So the feasible directions method relies much on line searches during which there is no need for sensitivity analysis. This is beneficial in our case and also improves the robustness.
Notice, though, that the algorithm is sensitive to different ranges of variables as they occur when you have variables of different dimensions. It is sometimes a good idea to formulate your design problem in intermediate variables that all vary, for instance, between 0 and 1 and then let your real variables be tied to the intermediate variables.
John
Thank you!
Is there in AnyScript a function like “time()” in C++, which show the time of Study working?
Hi Pavel,
Almost every object in an AnyScript model has a property called inside it. This is the time used by the object, and you can access thkis variable to know what the time is.
Best regards,
John