## For very large systems, direct solution methods (e.g.

*LU*decomposition) are slow and RE prone.## Often iterative methods much more efficient:

- Guess a trial solution
**x**0 - Compute a correction
**x**1 =**x**0 + **x** - Iterate procedure until convergence, i.e. |
**x**| < D

- Guess a trial solution
## e.g. Congugate gradient method for sparse systems (NRiC 2.7).