Re: How do I bisect unbounded intervals?
On 2012-01-13 03:13:24 +0100, Vincent Lefevre wrote:
> On 2012-01-13 03:11:28 +0100, Vincent Lefevre wrote:
> > > May I even be so bold as to suggest the "midpoint" of any
> > > interval [u,v] could be NaN if u and v are two adjacent Level 2
> > > datums, regardless if u and v are both finite or not? The
> > > rationale: this also avoids infinite recursion in naive B&B
> > > algorithms if midpoint of [u,v] is otherwise defined to be u or
> > > v.
> >
> > This may be a good idea. And this would mean that the following
> > property would hold at Level 2: u < midpoint([u,v]) < v.
>
> ... of course, when NaN doesn't occur.
Hmm... The case [u,u] has been forgotten. What would the spec be?
Can this happen in B&B algorithms?
--
Vincent Lefèvre <vincent@xxxxxxxxxx> - Web: <http://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <http://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)