Dmitrey, hi! Just to make it clear. What you are asking for is
ALGENCAN to call, at each iteration, to a user-provided subroutine.
Moreover, this subroutine should receive as parameters: the current
point, the objective function value and the sup-norm of the
constraints. Is it ok?
If this is what you are asking for, ok, we can do that. However, it
would imply in adding a new function to each ALGENCAN interface
(Python, C, Octave, R, Matlab, AMPL, CUTEr, etc). So, please, be
patient. We have a long list of pending improvements and/or
modifications.
Regarding a user-supplied stopping criterion subroutine, well it is a
little bit more delicate. ALGENCAN aims to satisfy feasibility,
optimality and complementarity at its final iterate. The tolerances
for those measurements being input parameters of ALGENCAN. Just users
with some experience may be able to replace the ALGENCAN stopping
criteria. Anyway, ok, we can add this to our list of pending tasks
too. (Again, please, be patient.)
Ernesto.
On Jun 16, 6:23 pm, dmitrey <dmitrey.kros...@scipy.org> wrote:
> hi,
> could you supply ALGENCAN 2.0 with callback function? (for iteration
> output)
> something similar to endp routine would be OK, but I would gladly have
> at least 3 callback values: current point, objfunc value and max
> residual.
> Also, it would be nice would it consume user-defined stop criterion.
|
