On my G5 iMac -
x=: 1000 1000 [EMAIL PROTECTED] 0 NB. random real matrix
6!:2 'y=: %. x' NB. invert it
13.7781
>./|, (=i.1000) - x +/ .* y NB. how accurate?
1.97126e_13
- joey
At 08:00 -0800 2007/01/15, Roger Hui wrote:
Regarding the limit on %. in J: Perhaps you should
get a bigger machine? I tried the following just
now on this machine, a run-of-the-mill Celeron
(2.8 GHz, 512 MB):
x=: 100 100 [EMAIL PROTECTED] 0 NB. random real matrix
6!:2 'y=: %. x' NB. invert it
0.0141401
>./|, (=i.100) - x +/ .* y NB. how accurate?
1.64729e_14
If numerical instability is a problem, there is
the option to use rational numbers, that is,
invert x: x instead of x . It'd be much more
expensive in space and time but %.x: x would find
THE inverse if there is one.
----- Original Message -----
From: Henning P. Jørgensen <[EMAIL PROTECTED]>
Date: Monday, January 15, 2007 7:26 am
Subject: [Jgeneral] Invert max and LAPACK
On my system it is possible to invert up to 87*87 matrices with J
and widows on a PC.
I believe excell goes to something like 64*64.
Of cause SAS and GAMS can go higher, but
Is it possible to get J to invert larger matrices? May be as Sparse?
As you may have guessed the LAPACK lab did not work on my PC.
I can download it to the add on area, but a call from the LAPACK
lab results in an error. Is there a solution to that?
Error in jijs_jctrl_fkey..
Best wishes
Med venlig hilsen
Henning P. Jørgensen
Lektor, Institut for Miljø- og Erhvervsøkonomi
Tlf. 6550 4187
Fax 6550 1091
Web http://www.sam.sdu.dk/ansat/hpj
Adr. Niels Bohrs Vej 9, 6700 Esbjerg
SYDDANSK UNIVERSITET
_______________________________________________________________
Niels Bohrs Vej 9-10 * 6700 * Esbjerg * Tlf. 6550 1000 * www.sdu.dk
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
