Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: dangerous motion 35 Kaucher arithmetic



Jürgen Wolff von Gudenberg wrote:
Dear Svetoslav,
   citing your last sentence
> The relation between usual interval arithmetic and Kaucher arithmetic
> is precisely the same. Kaucher (interval) arithmetic is for standard
(interval)
> arithmetic the same as real (number) arithmetic is for nonnegative
numbers
> arithmetic.

I conclude that Kaucher arithmetic needs or is defined with respect to
standard interval arithmetic, which we are going to standardize in P1788.
Hence we can , and should, define Kaucher arithmetic after we have
finished P1788
Juergen

On the one hand, I see this lends credibility to motion 35, since if P1788
does not act carefully right now, it may not be possible for the committe to
accomplish even that much without first making 1788 obsolete. And that
seems more "dangerous" to me, as it may completely undermine the credibility
of any work produced by P1788 in the meantime.

On the other hand, I agree with Svetoslav that putting Kaucher/modal
intervals at the end of a long list of "interval flavors" is rather
obfuscational to the real scientific point, which I think has been clarified
by the exchange between Svetoslav and Jeurgen.

Nate

P.S. I would remind people that there were originally similar claims that
the sky would fall if we included mid-rad to the standard, but P1788 solved
that problem really well by introducing explicit and implicit types. IMHO
P1788 needs the same attitude about Kaucher/modal intervals. Its amazing
what a little hard work and cooperation can do sometimes.




Am 01.07.2012 16:37, schrieb Svetoslav Markov:

Dear Juergen,
Dear prof Kulisch,

  To Juergen:  By:

" I see the danger that every system of interval arithmetic claims to be
1788 conforming, at least as a special flavor."

it seems you mean  the  "systems of interval arithmetic"   mentioned
in the letter  prof Kulisch: " Real intervals, set-based intervals,
standard
  intervals, classical intervals,  common intervals, conventional
  intervals, wrap around intervals, Kaucher and  modal intervals,
  and so on."

I have a question:

How can I  learn about all the enlisted types of intervals
with all the different  "flavours" between them? Which are
the remaining "so on" intervals ?

My comments.  In the above list I see only two main types of intervals:

  --- the usual intervals (under different names: "Real intervals,
set-based
intervals, standard intervals, classical intervals,  common intervals,
conventional intervals, wrap around intervals (?)", and

--- "Kaucher and  modal intervals", (which is the same)

Putting Kaucher/modal  intervals in such a long list of "intervals"
makes me think that something in your  letters  is not connected
just to science.

Juergen, you are an author of a paper based on  Kaucher arithmetic:

Wolff v. Gudenberg, J., Determination of Minimum Sets of the Set of
Zeros of a Function, Computing 24, 1980, 203–212.

and you know very well the unique role of Kaucher arithmetic.
What is the point of speaking about "every system of interval
arithmetic"?
Which "systems of interval arithmetic" do you have in mind? Which
are the systems that you can put on the same plane as Kaucher's one?

Kaucher interval arithmetic is the unique possible
group extension perserving inclusion - the main
  interval property. In the same way as real arithmetic is the
unique possible extension to the arithmetic over nonnegative
numbers perserving the distributive law.

Taking into account that many people in this forum do not know
well-enough this topic, makes me to conclude that  your mentioning the
  "many competing interval systems" is sort of obfuscation tactics,
aiming to
devaluate the special role of Kaucher arithmetic.

I  hope that this is not the case. I hope that a primary aim of this
forum is first to educate us in the various aspects of interval
arithmetic
and then to take correct decisions.

I urge the participants who know too little algebra to learn a bit more
about the relation between a monoid and a group, in particular between
(\R^+, +) and (\R, +) and to think why the IEEE standard mentions
real numbers (nonnegative plus negative), despite the fact that it
actually specifies operations on signed-magnitute reals, that is on
nonnegative numbers.

The relation between usual interval arithmetic and Kaucher arithmetic
is precisely the same. Kaucher (interval) arithmetic is for standard
(interval)
arithmetic the same as real (number) arithmetic is for nonnegative
numbers
arithmetic.

Best regards,
Svetoslav



On 30 Jun 2012 at 14:49, J"urgen Wolff von Gudenberg wrote:

Date sent:      Sat, 30 Jun 2012 14:49:56 +0200
From:           J"urgen Wolff von Gudenberg <wolff@xxxxxxxxxxxxxxx
WUERZBURG.DE>
To:             John Pryce <j.d.pryce@xxxxxxxxxxxx>,
        stds-1788 <stds-1788@xxxxxxxxxxxxxxxxx>
Subject:        dangerous motion 35

John, P1788
It is indeed an interesting task to analyze the relations of various
interval flavors, how they include each other, define commonly many
operations but also show contradictions in others.
This discussion should not be performed under the framework of P1788,
because    firsty  we do not have the time and secondly
I see the danger that every system of interval arithmetic claims to be
1788 conforming, at least as a special flavor.
I think P1788 will only be accepted if it is presented by a clear and
unique formulation of level 1 and a mapping to machine intervals in
level 2
furthermore a simple and reproducible specificaton of operations in
level 3 will enhance the dissemination

Juergen
--
       o           Prof. Dr. Juergen Wolff von Gudenberg,  Lehrstuhl
fuer
Informatik II
      / \          Universitaet Wuerzburg, Am Hubland, D-97074 Wuerzburg
InfoII o         Tel.: +49 931 / 31 86602
    / \  Uni       E-Mail: wolff@xxxxxxxxxxxxxxxxxxxxxxxxxxx
   o   o Wuerzburg




--
     o           Prof. Dr. Juergen Wolff von Gudenberg,  Lehrstuhl fuer
Informatik II
    / \          Universitaet Wuerzburg, Am Hubland, D-97074 Wuerzburg
InfoII o         Tel.: +49 931 / 31 86602
  / \  Uni       E-Mail: wolff@xxxxxxxxxxxxxxxxxxxxxxxxxxx
 o   o Wuerzburg