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

Re: Are there applications of the 7 comparison operators? -Resent



 Arnold Neumaier wrote in his mail below:
Other comparisons were also mentioned to be needed by other mails.

Could you please collect for reference all the uses mentioned in
one single mail?

This is my answer:
I did not collect these and I simply do not have the time to look for them right now. I was at the SCAN meeting last week and have to work through 150 mails now. I hope that somebody else has such a collection. I remember having seen something like this.

In the original proposal I had only four comparisons. By requests or suggestions during the discussion period in April I extended these to the seven that are now proposed. So this is a compromise. But I think it is a good compromise. Implemetation of comparisons is very simple in hardware as well as in software. So I really do not see a big problem. The relations are clearly defined and can easily be remembered.

Ulrich Kulisch

Am 03.10.2010 17:44, schrieb Arnold Neumaier:
Ulrich Kulisch wrote:
[in: Re: P1788: PLEASE VOTE: M0013.04 & M0020.01]

 Am 02.10.2010 18:17, schrieb Arnold Neumaier:
Corliss, George wrote:

Voting on Motion M0013.04 Comparison Relations ends on Friday, October 8.
Current count: Yes: 13; No: 0; Required for quorum: 37

Voting on Motion M0020.01 Comparison Relations ends on Friday, October 8.
Current count: Yes: 10; No: 2; Required for quorum: 37



The complexity pmomoted in these motions is not warranted by the applications. After having challenged the use of these relations,
not a single application was pointed out.

This is not correct. Applications using <= were mentionend in a mail by Juergen Wolff von Gudenberg.

The application to branch and bound is spurious:
For pure constraint satisfaction problems that are solved to completion,
the ordering used is completely immaterial, and using a stack no
ordering is needed at all.
But if the branch-and-bound cannot be completed for reasons of time or
space, or if used in a global optimization context, the <= ordering is
extremely inefficient, and dynamical orderings are needed.


Other comparisons were also mentioned to be needed by other mails.

Could you please collect for reference all the uses mentioned in
one single mail?


The <= relation is the connection link between the set definition of the arithmetic operations for intervals and the explicit formulas for computing the result of the operations by the bounds of the operands, not just for real intervals but also for real interval vectors and interval matrices and again for complex intervals and for complex interval vectors and matrices.

This does not yet constitute an algorithmic use. it is a theoretical
result, and for this it suffices that the comparison is defined on the
theoretical level. The standard, however, is about what needs to be
implemented.


For the <= relation compatibility relations hold between the algebraic structure and the order stucture similar to the real numbers.. For instance: 1. If a <= b ==> a + c = b + c for all c. 2. If a <= b ==> -b <= -a. 3. If 0 <= a <= b and c >= 0 ==> ac <= bc and ca <= cb. Similar for division.

Again, this is a theoretical result only.

I do not deny that the <= order has useful theoretical properties
that help in understanding the structure of intervals and interval
linear algebra.

But I haven't seen any algorithmic uses for these.


Thus my statement remains valid that the only comparison operations
useful in practice are
  disjoint,
  subset (containment),
  interior (containment in the topological interior).


Arnold Neumaier






--
Karlsruher Institut für Technologie (KIT)
Institut für Angewandte und Numerische Mathematik (IANM2)
D-76128 Karlsruhe, Germany
Prof. Ulrich Kulisch

Telefon: +49 721 608-2680
Fax: +49 721 608-6679
E-Mail: ulrich.kulisch@xxxxxxx
www.kit.edu
www.math.kit.edu/ianm2/~kulisch/

KIT - Universität des Landes Baden-Württemberg und nationales Großforschungszentrum in der Helmholtz-Gemeinschaft