A modified ODE-based algorithm for unconstrained optimization problems (Q2248961)

The authors propose a modified ODE-based trust region method for unconstrained optimization problems. At each iteration, a lower dimensional linear system is solved to obtain a trail step so as to avoid solving a trust region subproblem, and the approximation of the Hessian matrix to be positive definite or semidefinite is not required. The global convergence of the algorithm is established under certain conditions. The computational behavior of the proposed algorithm is illustrated by some numerical results for some testing problems.