Implicit and adaptive inverse preconditioned gradient methods for nonlinear problems (Q2387319)

The authors extend recently developed numerical schemes for approximating the inverse of a given matrix using an ordinary differential equation (ODE) model to solve nonlinear unconstrained optimization problems \(\min f(x)\) subject to \(x\in\mathbb{R}^n\) with a continuously differentiable objective function. The main idea is to develop an automatic and implicit scheme to approximate directly the preconditioned search direction at every iteration without an a priori knowledge of the Hessian of the objective function. The procedure is appropriate mainly for large-scale problems for which the Hessian of \(f\) is either unavailable or requires a prohibitive amount of storage and computational cost. The new scheme avoids the cost of any calculations involving matrices and makes possible to obtain the Newton direction by improving the accuracy in the ODE solver associated with the implicit scheme. Extensive numerical results on some well-known test problems are according to authors' opinion encouraging.