Isn't the log of negative numbers indefined?
This is a problem:
%1&o.+0
_
%1&o.-0
_
This is nice!
%1&o.%_
_
%1&o.%__
__
The csc is very small for negative numbers close to zero and very large for
very small positive numbers.
Linda
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Bo Jacoby
Sennt: Thursday, January 17, 2013 3:37 AM
To: [email protected]
Subject: Re: [Jprogramming] Hermitian from triangular
Henry, How is negative zero different from positive zero when taking the
log?
^.%__ NB. log -0
__
^.%_ NB. log +0
__
- Bo
>________________________________
> Fra: Henry Rich <[email protected]>
>Til: [email protected]
>Sendt: 0:38 torsdag den 17. januar 2013
>Emne: Re: [Jprogramming] Hermitian from triangular
>
>Negative zero makes sense as a last vestige of gradual underflow; and
anyway, it's well-behaved: it looks like 0 except when you take the log,
reciprocal, or square root. In any normal computation, it goes away. In
contrast, NaN messes up anything it touches.
>
>I think we've had negative 0 in J forever. If NaN is a data virus, -0 is a
virus that has been inserted into our DNA.
>
>Henry Rich
>
>On 1/16/2013 4:45 PM, Raul Miller wrote:
>> On Wed, Jan 16, 2013 at 4:35 PM, Henry Rich <[email protected]> wrote:
>>> Negative zero isn't a bug, it's a feature that numerical types,
>>> especially William Kahan, wanted to get into IEEE-754 to help out
>>> some things. I'm not expert enough to explain.
>>
>> Something similar could be said about NaN.
>>
>----------------------------------------------------------------------
>For information about J forums see http://www.jsoftware.com/forums.htm
>
>
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
