= Stationary iterative method for a linear system
A <stationary iterative method for a linear system> uses the same <iteration matrix> at every step:
$$
x^{(k+1)}=Hx^{(k)}+v.
$$
For a specified <fixed point> $x_*$ satisfying $x_*=Hx_*+v$, the <forward error> obeys $e^{(k)}=H^ke^{(0)}$. Convergence to that same $x_*$ for every starting <vector> is equivalent to $\rho(H)<1$, as the <convergence criterion for a stationary iteration> proves. Merely having a convergent sequence for every start is weaker: $H=I$, $v=0$ leaves each starting value unchanged.
For a <matrix splitting> $A=M-N$ with $M$ invertible, choose $H=M^{-1}N$ and $v=M^{-1}b$. Then every solution of $Ax=b$ is a <fixed point>. The choices of $M$ produce the <Jacobi method>, <Gauss-Seidel method> and <successive over-relaxation>. Fixed coefficients make the error behavior analyzable once, while the <vector> $b$ determines the desired solution.
Back to article page