On 8 Sep 2009, at 19:23, Arnold Neumaier wrote:
The NaI with payload is essentially equivalent to propagating only
the decoration while throwing away the interval. This saves space
but wastes information. In some cases, it may lead to a saving.
But it leads to a loss in all cases where both the interval and the
decoration contain useful information (e.g. in any interval Newton
method that wants to verify existence and uniqueness).
I seem to be suffering mental exhaustion with all the ideas flying
round, so I would appreciate a summary of the interval Newton method
(esp. the multivariate case) showing how and where both the interval and
the decoration are needed. It sounds as if it falls into the category I
named B3 in my email of 8th Sept.