Rescaled multilevel least-change almost secant methods

In this article the theory of local convergence is developed. One of the extensions consists in that the approximations to the Jacobian matrix have the same properties as the same matrix has in the solution. The example shows that this assumption may lead to simpler algorithms. The paper discusses several rescaled multilevel least-change updates for which local g-superlinear convergence is proved. The theory may be applied to a wider class of methods because every secant algorithm may be treated as a rescaled least-change method.