
Boost : 
From: Matt Sutton (matt.sutton_at_[hidden])
Date: 20021025 10:11:19
> Original Message
> From: Martin Weiser [mailto:weiser_at_[hidden]]
> Sent: Friday, October 25, 2002 1:05 AM
>
> On Freitag, 25. Oktober 2002 08:04, Daryle Walker wrote:
>
> > There are at least four ways to address the domain values:
> >
> > 1. initial x, final x, and the number of intervals between them
> > 2. initial x, final x, and the interval size
> > A. note that the interval size may not fit evenly
>
> Probably a common case if evenly spaced data is available. 1
> seems to be a
> more or less trivial special case of 2.
>
> > 3. initial x, final x, and letting the algorithm decide
> intermediate
> > x's A. this is for an adaptive algorithm
> > B. requires the y's to be given by a function/functor
>
> Probably the most apropriate case if the data is indeed given by a
> function/functor. Also allows for the most sophisticated, efficient,
> robust, and accurate algorithms. That's the case I'm most
> interested in.
[snip]
At a minimum, we would need to supply an error criterion for case 3 as well,
with perhaps a suitable default value. From what I remember, exactly how
you compute the error within different algorithms can be a small bone of
contention among the purists. I don't know if delagating error computations
to a policy class such that users could supply specialized methods would be
overkill or not.
3. initial x, final x, error
Matt
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk