On Wed, May 27, 2009 at 1:49 PM,  <[email protected]> wrote:
> It is known that if you approximate a number x with continuous
> fraction a/b you can say that x is irrational if
>  ABS (a/b)* ABS (x-a/b)/x > 0,0033
> else you have a rational number.

This does not make sense, to me, in the context of
computer arithmetic.

First off, you have expressed this in terms of x,
which means you must know the exact value
of x (or have a suitably solid theoretical framework
for representing x) before you can use this
expression.

Second, your 0.0033 epsilon value seems completely
arbitrary, but

   2 ^. 0.0033
_8.24332

it's something near 8 bits of mantissa.  And IEEE floating
point notation represents some numbers with less
than 8 bits of mantissa so some rational numbers
represented using floating point could be trivially
classified as rational  Alternatively, this mechanism
depends on the quality of the approximation a/b (or,
using J's notation, of a%b).

That said, J's arithmetic can only treat a subset of
the rational numbers.

-- 
Raul
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to