Re: Successive Over-Relaxation ... what is wrong? Improvements?

[Ed Meyer]

On Sun, Nov 4, 2012 at 11:31 AM, Ed Meyer <address@hidden> wrote:

On Sat, Nov 3, 2012 at 6:12 PM, Joza <address@hidden> wrote:

Yea....

I've tried a few values for the relaxation parameter and the matrix size ...

it seems that either the norm converges to a constant, or it keeps
oscillating around a certain value.

As far as I am aware, I've followed the implementation as per textbook - I
can't see anything obviously wrong with it. Any idea what could be causing
this?



--
View this message in context: http://octave.1599824.n4.nabble.com/Successive-Over-Relaxation-what-is-wrong-Improvements-tp4646041p4646059.html

Sent from the Octave - General mailing list archive at Nabble.com.
_______________________________________________
Help-octave mailing list
address@hidden
https://mailman.cae.wisc.edu/listinfo/help-octave

It works for me if I use the formula for the overrelaxation factor
  w = 2/(1 + pi/n)
that is usually recommended

--
Ed Meyer

I would also point out the the example you are using is not diagonally dominant,
which is usually a requirement for convergence


-- 
Ed Meyer