New abstraction : at each iteration, you give it the stats and it returns the number of samples you need to do for this iteration and whether you need to do cuts or just estimate the upper bound.
It would make it possible to use sophisticated scheduling without the sddp algorithm needing to care about it.
What do you think ? Do you have any idea for a good name (maybe not a bird name here :-P) ?
New abstraction : at each iteration, you give it the stats and it returns the number of samples you need to do for this iteration and whether you need to do cuts or just estimate the upper bound. It would make it possible to use sophisticated scheduling without the sddp algorithm needing to care about it. What do you think ? Do you have any idea for a good name (maybe not a bird name here :-P) ?