Statistics preprint
Author Message
Statistics preprint

Gardyloo!
I have just posted some statistics routines to
http://www.*-*-*.com/ ~lenz/forth.htm
I hope some of you find them of interest.
This is code I plan to talk about at the upcoming
Forth Technology Symposium.  If you spot any errors
in what is posted I would appreciate hearing about them
soon enough to not be embarrassed by them on June 20th.
As always, test results and general suggestions are also welcome.

There also may be some philosophical issues in the way I
approached the problem.
-LenZ-

Wed, 28 Nov 2001 03:00:00 GMT
Statistics preprint
Help!  As usual (some day I really *have* to look up Paranoids
Anonymous, but can I trust them?) I am having second thoughts
about what I wrote.  The problem is ANS Forth's (increasingly
annoying) lack of any reasonable method of detecting overflow.

The question at hand: in the DOES> part of "collect" is there
any combination of input that will cause the calculation of an
of overflow?  I don't think so, but I am having trouble
proving it.  I won't be really happy until I either have
a proof it can't or an example set of input that makes it happen.
If there is going to be a problem I think it will come with a
large negative input value and an offset subtraction that will
come out on the number circle as a small positive result.

Can any of you math mavens in web land help me out?
-LenZ-

Quote:

> Gardyloo!
> I have just posted some statistics routines to
> http://www.flash.net/~lenz/forth.htm
> I hope some of you find them of interest.
> This is code I plan to talk about at the upcoming
> Forth Technology Symposium.  If you spot any errors
> in what is posted I would appreciate hearing about them
> soon enough to not be embarrassed by them on June 20th.
> As always, test results and general suggestions are also welcome.

> There also may be some philosophical issues in the way I
> approached the problem.
>   -LenZ-

Thu, 29 Nov 2001 03:00:00 GMT
Statistics preprint
Just cause you're paranoid doesn't mean they *aren't* out to getcha!
Managed to answer my own question :-(
Trying to collect statistics with the smallest negative integer
followed by the largest positive integer won't work.
-LenZ-
Quote:

> Help!  As usual (some day I really *have* to look up Paranoids
> Anonymous, but can I trust them?) I am having second thoughts
> about what I wrote.  The problem is ANS Forth's (increasingly
> annoying) lack of any reasonable method of detecting overflow.

> The question at hand: in the DOES> part of "collect" is there
> any combination of input that will cause the calculation of an
> adjusted input value to come up with the wrong answer because
> of overflow?  I don't think so, but I am having trouble
> proving it.  I won't be really happy until I either have
> a proof it can't or an example set of input that makes it happen.
> If there is going to be a problem I think it will come with a
> large negative input value and an offset subtraction that will
> come out on the number circle as a small positive result.

> Can any of you math mavens in web land help me out?
>    -LenZ-

Thu, 29 Nov 2001 03:00:00 GMT
Statistics preprint

Quote:

> Just cause you're paranoid doesn't mean they *aren't* out to getcha!
> Managed to answer my own question :-(
> Trying to collect statistics with the smallest negative integer
> followed by the largest positive integer won't work.
>   -LenZ-

And the next round is now complete and loaded to
http://www.flash.net/~lenz/forth.htm

The wolves may now attack.

Comments, criticisms, bug reports (especially bug reports)
welcome.

-LenZ-

Sat, 01 Dec 2001 03:00:00 GMT

 Page 1 of 1 [ 4 post ]

Relevant Pages