We investigate a constrained version of simultaneous iterative reconstruction techniques (SIRT) from the general viewpoint of projected gradient methods. This connection enable us to assess the computational merit of this algorithm class. We borrow a leaf from numerical optimization to cope with the slow convergence of projected gradient methods and propose an acceleration procedure based on th...

In this paper we propose a nonmonotone approach to recurrent neural networks training for temporal sequence processing applications. This approach allows learning performance to deteriorate in some iterations, nevertheless the network’s performance is improved over time. A self-scaling BFGS is equipped with an adaptive nonmonotone technique that employs approximations of the Lipschitz constant ...

The primary purpose of this research is the analysis of nonmonotone optimization algorithms to which standard convergence analysis techniques do not apply. We consider methods that are inherently nonmonotone, as well as nonmono-tonicity induced by data perturbations or inexact subproblem solution. One of the principal applications of our results is the analysis of gradient-type methods that pro...

We study and analyze a nonmonotone globally convergent method for minimization on closed sets. This method is based on the ideas from trust-region and Levenberg-Marquardt methods. Thus, the subproblems consists in minimizing a quadratic model of the objective function subject to a given constraint set. We incorporate concepts of bidiagonalization and calculation of the SVD “with inaccuracy” to ...