Re: Q: minimize

Warwick Harvey (wh@icparc.ic.ac.uk) wrote:

: Apparently FD's minimize/2 can be a bit late in imposing the new cost
: bound.  The general problem here is that if you impose a new bound after
: finding a solution, backtracking occurs, immediately retracting the new
: bound.  What one really wants is a way to have the bound re-imposed after
: any failure, simulating it being applied in a more global context.
: Facilities for doing this were not available in ECLiPSe until quite
: recently, so back when this library was written other (not as good)
: techniques had to be used instead.

I see.
 
: Anyway, I suggest you try using the branch_and_bound library.  The
: minimize/2 there seems to do it right (using the new facilities), and I'm

I've started tu use bb_min which is also quite nice.

: led to believe that's the preferred library to use for this stuff.  Probably
: FD's routines were not updated when the new techniques became available
: since we are planning to phase that library out.

Thanks for the explantion and info.

Regards,



Marc van Dongen