xuy
Documentation on apply_new
I've done some searching but haven't found an explanation of what the apply_new function does. I started with
def apply_new(x):
return True
as the examples show. However I have found that returning false improves performance a bit but I'm not sure why.
I believe it's a callback function to indicate whether the Hessian of the Lagrangian is user-provided. See the start page https://github.com/xuy/pyipopt for a short mention of it. I have found little mention of it elsewhere, nor how to change away from the default return value "True".
The readme that you reference says that if the functions "eval_h" and "apply_new" are provided then hessian approximation isn't used, but it doesn't have any mention of the effect of "apply_new" return value .
In my experience setting apply_new to false still didn't use Hessian approximation and it actually caused my problem to converge faster. I wonder if perhaps apply_new is used to indicate whether or not to re-evaluate the Hessian. In my case the Hessian is constant so it would make sense that there is no need to reevaluate it.