Computable error bounds and estimates for the conjugate gradient method (Q5934354)

It is the aim to get error bounds for an approximate solution to linear systems of equations, e.g. for the iterates of a \(cg\)-iteration. When this is done for the energy norm, only a term of the form \(b^T A^{-1} b\) is not explicitly known. In principle one could use a \(cg\)-iteration for this purpose, but then one would rather spend the effort for performing further steps of the original iteration. Now, a lower bound can be obtained from a Gaussian quadrature rule and an upper bound from an anti-Gauss quadrature.