Um metodo Newton-Inexato com estrategia hibrida para globalização / A Newton-Inexact algorithm with a hybrid method for globalization
AUTOR(ES)
Rodolfo Gotardi Begiato
DATA DE PUBLICAÇÃO
2007
RESUMO
The main objective of this work is to propose a hybrid globalization strategie for inexact-Newton method. Globalization strategies are based on line search or trust region procedures. In this work, we choose a hybrid strategy which involves a cycle of line search and a variation of Powell dogleg trust region. For solving the linear systems we chose the GMRES method with restarts and to avoid the calculation of Jacobian matrices we used a matrix-free strategie. The numerical performance of the algorithms was evaluated by means a set of academic problems and a set of nonlinear systems of boundary value problem discretization. These results showed the good performance of hybrid globalization strategy
ASSUNTO(S)
sistemas não-lineares metodos interativos (matematica) otimização matematica nonlinear systems iterative methods mathematical optimization
ACESSO AO ARTIGO
http://libdigi.unicamp.br/document/?code=vtls000414653Documentos Relacionados
- An inexact subgradient algorithm for Equilibrium Problems
- A globally convergent Newton-GMRES method with a new choice for the forcing term and some stragies to improve GMRES(m)
- AN HYBRID METHOD FOR PHOTOELASTIC ANALYSIS OF ORTHOTROPIC COMPOSITE MATERIALS
- Convergência local do método de Newton inexato e suas variações do ponto de vista do princípio majorante de kantorovich
- Convergência local do método de Newton inexato e suas variações do ponto de vista do princípio majorante de kantorovich