Evidently you don't consider that 

   1 (1 teq) 1e9
1

is not a "meaningful result".  Even casual further experimentation
produces:

   1 (1 teq) 0 1 2 3 4 5 6
1 1 1 1 1 1 1
   1 teq"0/~ 10 [EMAIL PROTECTED] 1e6
1 1 1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 1

is that meaningless enough for you?


 
----- Original Message ----- 
From: "Oleg Kobchenko" <[EMAIL PROTECTED]>
To: "Programming forum" <[email protected]>
Sent: Tuesday, May 30, 2006 10:33 AM
Subject: Re: [Jprogramming] Equal tolerance fit conjunction -- again

It says "tolerances greater than or equal to 1 do 
not give meaningful results". And then it gives
one example of a "bad" case
   1 (1 teq) 1e9
1

I give other examples that "work" for 1<t<2.

So what is a "meaningful" result?

What should we require of t in order to 
restrict its domain?

If p is the one with the larger magnitude, than
the domain of q (with less than or equal marnitude)
is (-|p)..|p (or circle with radius |p at origin for complex).
The task of identification is to classify in two groups: 
equal or not equal. Domain of t will determine the quality 
of how well it will be able to classify.
 - t should be non-negative real, because the radius is non-negative
 - if t is 0, the radius is 0 so it will identify only one q = p
 - if t=1 it will identify exactly half of domain of q for reals
   (or intersection of cicles of radius |p and origins at 0 and p)
 - if t>:2 it will identify all q

So why should 1 as a break point be preferred over 2?


--- Roger Hui <[EMAIL PROTECTED]> wrote:

> The fact that t must be less than 1 is demonstrated
> in the tolerant comparison essay.
> http://www.jsoftware.com/jwiki/Essays/Tolerant_Comparison
> Since I had already cited the essay in this thread
> and you yourself had also cited it, I assume you have
> read it.  Then I don't understand why you said what you
> did in (*).  The essay states and demonstrates that the
> tolerance t must be less than 1.  Why do you then bring
> up cases where t is greater than 1?
> 
> 
> 
> ----- Original Message ----- 
> From: "Oleg Kobchenko" <[EMAIL PROTECTED]>
>   *) it's not totally clear how it is a consequence
> that t must be less than 1. Something is happening
> between 1 and 2 as well:
>    _1 (1.9 eq) 1
> 0
>    _1 (2 eq) 1
> 1
>    _1 (1.4 eq) 0.5
> 0
>    _1 (1.5 eq) 0.5
> 1


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

Reply via email to