Resumo
In 2017, a method for solving the trust-region subproblem using generalized eigenvalues was rediscovered and improved. In 1989, when the original method was proposed, it presented a poor performance, which was caused by the low quality of eigensolvers available at that time. In this work we explore some geometric characteristics of this method.

Este trabalho está licenciado sob uma licença Creative Commons Attribution 4.0 International License.
Copyright (c) 2019 Jean Carlos Aparecido Medeiros, Sandra Augusta Santos