PCG methods applied to a system of nonlinear equations (Q1184108): Difference between revisions

Let \(A\) be an \(m\times m\) symmetric positive definite block tridiagonal \(M\)-matrix and \(G\) a bounded continuous function from \(\mathbb{R}^ m\) to \(\mathbb{R}^ m\). The authors study the approximate solution of the equation \(F(x):=Ax+G(x)=0\) by means of the Newton-like method \(B(x_ m)(x_ n- x_{n+1})=F(x_ n)\), \(n\in\mathbb{N}_ 0\), where the updating matrices \(B(x_ n)\) are chosen \(B(x_ n):=A+\phi(x_ n)\) with a certain function \(\phi\) such that \(B(x_ n)\) satisfies the quasi-Newton equations \(B(x_ n)(x_ n-x_{n-1})=F(x_ n)-F(x_{n-1})\), \(n\in\mathbb{N}\). The linear system is solved using the preconditioned conjugate gradient (PCG) method. The preconditioner is based on the structure of \(A\) and fixed for all \(n\in\mathbb{N}\). The PCG method to solve the linear system at each step of the Newton-like iteration is discussed for various preconditioners, the condition numbers of which are estimated and the computing times are compared for a numerical example.