Paul McGuire wrote: > On Jan 24, 1:47 pm, Robert Kern <[EMAIL PROTECTED]> wrote: >> Paul McGuire wrote: >>> And the purpose/motivation for "reimplementing it better" would be >>> what, exactly? So I can charge double for it? >> So you can have accurate results, and you get a good linear solver out of the >> process. The method you use is bad in terms of accuracy as well as >> efficiency. > > Dang, I thought I was testing the results sufficiently! What is the > accuracy problem? In my test cases, I've randomly created test > matrices, inverted, then multiplied, then compared to the identity > matrix, with the only failures being when I start with a singular > matrix, which shouldn't invert anyway.
Ill-conditioned matrices. You should grab a copy of _Matrix Computations_ by Gene H. Golub and Charles F. Van Loan. For example, try the Hilbert matrix n=6. H_ij = 1 / (i + j - 1) http://en.wikipedia.org/wiki/Hilbert_matrix While all numerical solvers have issues with ill-conditioned matrices, your method runs into them faster. -- Robert Kern "I have come to believe that the whole world is an enigma, a harmless enigma that is made terrible by our own mad attempt to interpret it as though it had an underlying truth." -- Umberto Eco -- http://mail.python.org/mailman/listinfo/python-list
