Gene, you may recall that back in the early 1980's you sent me an APL  
model for Quad-Divide, which used Househoulder Transforms for QR  
Decomposition to produce an upper triangular matrix as you described.   
A key aspect of this algorithm is that it was stable for non- 
rectangular matrices also (Least Squares).

I may still have the model at home, but I would think that Roger is  
the best one to answer this question for %. as he may have implemented  
a different model anyway.

Regards, Rob Hodgkinson

On 23/08/2009, at 3:18 PM, Eugene McDonnell wrote:

>
> On Jul 6, 2007, at 12:48 AM, Damien wrote:
>
>> Hello,
>>
>> I was wondering what kind of algorithm is used within the %.  
>> operator,
>> because it is very powerful !! Does someone have any idea ?
>> Besides, what is the best algorithm to invert matrix for you ? I
>> mean if you
>> had to implement only one algorithm for all the types of matrices,
>> which one
>> would you choose ?
>>
>> Thanks.
>> Damien.
>
> My dim recollection is that I triangularized Matrix M, giving Matrix
> T, and then the inverse is easy to compute. From that I went backwards
> and got the inverse.
>
> I've just finished looking in my files on Matrix problems, but was
> unable to find anything; very vexing.
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm

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

Reply via email to