Hi Christian,
Thanks for your reply. I haven't been able to look again at this issue
until now, so sorry for the delay...
On 13/09/2013 12:20, Christian Schulte wrote:
As for making sure that the same solution is not found again there is some
support but not really the right one. For integer variable arrays x and y
(also for Boolean arrays) you can post by
rel(home, x, IRT_NQ, y);
cannot be the same, that is there is at least one i for which x_i is
different from y_i.
Am I correct that for the above, x would be the variables that are being
labelled by the search (i.e. the variables passed to branch()), and y is
the solution values for these variables (x and y can be swapped, of
course)? Also, this rel constraint should be posted in master()?
I am not quite sure how to pass y to the master() function -- I assume
you need to somehow remember it (as with the no-goods).
If this is not easy to do, then I think an alternative behaviour I can
implement is to not allow the user to ask for an alternative solution if
restart is used with DFS. Am I correct in thinking that restart is
probably not that useful if you want to search for multiple solutions of
a problem, because the search (unless you use some randomisation, as you
mentioned) will always drive you close to solutions you already found,
before exploring other parts of the search space.
Cheers,
Kish
_______________________________________________
Gecode users mailing list
users@gecode.org
https://www.gecode.org/mailman/listinfo/gecode-users
