Re: [Reliable Computing] abs[x] for intervals? (reason for a standard)
Corliss, George a écrit :
[ Some good remarks here. ]
I hate to be elitist, but I think intervals are only usable by experts.
But, isn't floating-point for experts as well anyway ?
We are basically going to have IA "compete" with FP (typically as
a way to compute approximations of reals). Users will have to choose
between FP and IA for a given problem, with many arguments to form
their decision like ease of use, ease of provability, teachability
of the corresponding methods, ressource efficiency...
And hopefully IA will end up being *less* "for experts" than FP :
it might provide a less steep learning curve for programming with
reals (in a certified way), because it computes more information
which can be queried simply.
So, I think it could have a chance to become the "default for beginners"
at some point, or at least be considered for much more general use.
All this might require to devise guidelines for when to best use FP or IA.
And this is linked (both ways) with fast common hardware support, of course.
--
Sylvain Pion
INRIA Sophia-Antipolis
Geometrica Project-Team
CGAL, http://cgal.org/