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

Re: Re-submission of motion 5: multiple-format arithmetic.



John
in your last eply to the Kulisch-Einarsson motion  5 you wrote
". Then, your current p4 stuff is more or less saying the same thing as the definitions in my (under revision) multi-format motion, except you only mention the 4 basic ops, and I cover general multivariate point-functions, which _are_ necessary. So it would be nice for us to converge closer."
and now you again deny this in section R4.


John Pryce schrieb:
P1788 members

R4. Overall Aims.
-------------

Ramon Moore's Basic Theorem of Interval Arithmetic (BTIA) is fundamental to interval computation. Roughly, it says that if E is an explicit real expression defining a real function f(x_1, ..., x_n), then evaluating E "in interval mode" over any interval inputs (xx_1, ..., xx_n) is guaranteed to give an enclosure of the range of f over those inputs.

This motion has two linked aims. Aim 1 is to define a framework for finite precision interval arithmetic in multiple formats. In this framework the BTIA is easily proved -- indeed almost obvious. The finite precision interval operations are defined to operate on a set of abstract objects that happens -- except for Not an Interval, NaI -- to be a finite subset of the infinite set (IR). The definition of operations is independent of any representation that may be chosen for intervals.

The framework should support mixing intervals of different formats in arithmetic expressions, as well as potential compile-time "exact denotations" of intervals such as pi + [-0.1,0.1]. However, this motion does not commit P1788 to any stance on such issues.

[Note: To simplify wording I am assuming P1788 defines an NaI object. Personally I support NaI; editorially I have no preference; and NaI is peripheral to the argument.]

This is exactly analogous to how 754 defines the finite precision point operations, on a set of abstract objects that happens -- except for -0, +0 and NaN -- to be a finite subset of the extended real numbers.

The abstract objects and operations form interval level 2. To see its utility, ask yourself the question "How to prove the BTIA, starting from a procedural definition of the arithmetic operations as given, for instance, in the Einarsson-Kulisch position paper presently (May 2009) under discussion as Motion 5?" Answer: you cannot, _except_ by passing through a stage equivalent, or closely related, to the abstract objects and operations defined here. (This is no criticism of these definitions: they have a different purpose.)

--
=======
      o          Prof. Dr. J. Wolff v. Gudenberg,  Informatik 2
     / \         Univ. Wuerzburg,  Am Hubland,   D-97074 Wuerzburg
 info2 o        Tel.: +49 931 / 31-86602  Fax: +49 931 / 888-6603
   / \  Uni             e-mail: wolff@xxxxxxxxxxxxxxxxxxxxxxxxxxx
  o   o Wuerzburg         http://www2.informatik.uni-wuerzburg.de/