Re: Branch & bound for not everywhere defiend constraints
The various examples where domain violations are mishandled if the
wrong choice is made with regard to returning or not returning NaI
demonstrate that the choice is needed. The problem now is that
this requires somebody to make the choice, i.e. understand what is
going on. In other words, the programmer has to THINK (oh my, how
old-fashioned can you get).
I am not familiar enough with the techniques discussed in detail
by the participants in this discussion to evaluate the difficulties
that this entails.
Arnold Neumaier's approach of computing according to the containment
rule, but noting domain violations on the side, permits the choice
to be deferred. So does the method of tagging intermediate results
with flags. Both of these have implementation and performance issues.
The two-opcode approach requires a choice to be made ahead of time,
which may be more difficult: less information, so sometimes BOTH
methods may have to be used to permit choosing after the fact. So
no global decision appears to be able to avoid superfluous work all
the time!
Michel.
---Sent: 2009-09-03 10:57:12 UTC