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

[c.]

On 4 Nov 2012, at 14:05, Joza wrote:

> Yea, for A = 10x10, for example, here is the result, with relaxation
> parameter = 1.4:
> 
> ...
> x: 4.999998     iters: 710      norm: 0.0000010676089660 
> x: 4.999998     iters: 711      norm: 0.0000003683371855 
> x: 4.999998     iters: 712      norm: 0.0000006345346101 
> x: 4.999998     iters: 713      norm: 0.0000007962736616 
> x: 4.999998     iters: 714      norm: 0.0000008603756717 
> x: 4.999998     iters: 715      norm: 0.0000008603756717 
> x: 4.999998     iters: 716      norm: 0.0000008603756717 
> x: 4.999998     iters: 717      norm: 0.0000008603756717 
> x: 4.999998     iters: 718      norm: 0.0000008603756717 
> x: 4.999998     iters: 719      norm: 0.0000008603756717 
> x: 4.999998     iters: 720      norm: 0.0000008603756717 
> 
> ans =
> 
>    5.0000
>    9.0000
>   12.0000
>   14.0000
>   15.0000
>   15.0000
>   14.0000
>   12.0000
>    9.0000
>    5.0000
> 
> So it seems to work here?
> 

not really, it actually seems to be stagnating.

I just tried a very simple implementation of the algorithm wiyh your parameters
and it converges within a tolerance of 1e-15 in about 170 iterations.

c.