I attempted to use one of the math/mt routines (geev) but get a value error
on "trsmu1x" which  does not appear to be defined anywhere.

There is a Jacobi solver in ~addons/math/misc/jacobi that appears to work
for a symmetric argument.  Also, there's the code from Donald McIntyre's
article on Jacobi's method -
http://www.jsoftware.com/jwiki/DevonMcCormick/JacobiByMcIntyre - which I
modernized to run in contemporary J - that appears to give an answer when
the other version of "jacobi" does not (i.e. for a non-symmetric matrix)
but I don't know how good this is.  I can use this latter to duplicate the
results shown here -
http://algebra.math.ust.hk/eigen/01_definition/lecture2.shtml - for some
simple, small problems.


On Mon, Apr 8, 2013 at 3:17 AM, Kip Murray <[email protected]> wrote:

> Igor Zhuravlof provides j routines that model LAPACK routines for
> eigenvalues and eigenvectors.  See "matrix toolbox"
>
> ~addons/math/mt
>
> with contents summarized in
>
> ~addons/math/mt/mt.ijs
>
> I have not tried them but would expect them to run in j701JHS.
>
> --Kip Murray
>
>
> On 4/8/2013 12:45 AM, Michal D. wrote:
>
>> Hi All,
>>
>> I have been trying to get J to compute the eigen values & vectors of a
>> matrix.  I've come across
>> http://www.jsoftware.com/**jwiki/Essays/Eigenvaluesbut<http://www.jsoftware.com/jwiki/Essays/Eigenvaluesbut>none
>>  of the
>> algorithms there seem stable on my input matrix and the
>> lapack package doesn't seem to work on 64-bit J.  I laboriously exported
>> the matrix to a .csv and then hacked up a C++ program that computes the
>> eigen values using eigen (
>> http://eigen.tuxfamily.org/**index.php?title=Main_Page<http://eigen.tuxfamily.org/index.php?title=Main_Page>)
>> which had no problem
>> with computing the eigen values.
>>
>> What is the current state-of-the-art in eigen values in J?  Should I just
>> downgrade to 32-bit J and use the lapack package?
>>
>> Cheers,
>>
>> Mike
>>
>>
>> Ps. There is nothing I like better than generating random input data using
>> J.  What took pages of Haskell and TCL was boiled down to 7 clear lines of
>> J.  WOW =)
>> ------------------------------**------------------------------**
>> ----------
>> For information about J forums see 
>> http://www.jsoftware.com/**forums.htm<http://www.jsoftware.com/forums.htm>
>>
>>  ------------------------------**------------------------------**
> ----------
> For information about J forums see 
> http://www.jsoftware.com/**forums.htm<http://www.jsoftware.com/forums.htm>
>



-- 
Devon McCormick, CFA
^me^ at acm.
org is my
preferred e-mail
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to