Some convergence results related to the JOR iterative method for symmetric, positive-definite matrices (Q1186985): Difference between revisions

The author considers the modified Jacobi iterative method (JOR) for solving a system of linear algebraic equations \(Ax=b\) with a positive definite matrix \(A\). If \(A\) is decomposed as \(A=D-N\), where \(D=\text{diag }A\), then splitting of the form \(A=\alpha D-\{N-(1-\alpha)D\}=D_ 1-N_ 1\) is used for a construction of \(JOR\)-scheme of the form \(D_ 1x^{n+1}=b+N_ 1x^ n\). The author proves that this scheme converges for all symmetric positive definite matrices when \(\alpha > n/2\), where \(n\) is the dimension of \(A\) and for the optimum \(\alpha_{opt}\) the well known formula \(\alpha_{opt}=(\lambda_{max}(D^{-1}A)+\lambda_{min}(D^{- 1}A))/2\) is given.