• A NEW SELF-SCALING VARIABLE METRIC ALGORITHM FOR NONLINEAR OPTIMIZATION

Abstract

In this paper a new self-scaling Variable Metric (VM) method for solving a number of nonlinear unconstrained optimization problems is proposed. In this work we have suggested a new formula for the VM-update with a new Quasi-Newton (QN) like condition used for designing this update. Numerical experiments indicate that this new self-scaling method is effective and superior to both BFGS and Biggs VM-methods, with respect to the number of function evaluations (NOF) and the number of iterations (NOI).