A new trust region method for unconstrained optimization (Q2479356)

A new trust region method for unconstrained optimization problems and some convergence properties under some mild conditions are developed and presented. At each iteration the new trust region method generates a suitable initial trust region automatically based on the current iterative information. The global convergence and local superlinear and quadratic convergence rate are investigated. The implementation of the proposed method is resulting to numerical data that proof that the new trust region method is available and efficient in practical computation.