Direct secant updates of sparse matrix factors
Technical Report
·
OSTI ID:6255404
Quasi-Newton methods for the solution of large sparse systems of non-linear equations suffer two disadvantages versus quasi-Newton methods for small systems. They require a matrix factorization on each iterate and also need to maintain in storage both the Jacobian approximation and its factors at the same time. Two updating techniques, which are applied directly to the factors of the matrix approximation and which attempt to remove these two disadvantages, are discussed here. A theoretical analysis of local Q-superlinear convergence is presented.
- Research Organization:
- Stanford Univ., CA (USA). Systems Optimization Lab.
- DOE Contract Number:
- AT03-76ER72018
- OSTI ID:
- 6255404
- Report Number(s):
- SOL-84-5; ON: DE85004008
- Country of Publication:
- United States
- Language:
- English
Similar Records
Sparse quasi-Newton methods and the continuation problem
Hybrid algorithm for solving sparse nonlinear systems of equations
Performance analysis of nonlinear equation solution techniques for chemical process design
Technical Report
·
Sat Jun 01 00:00:00 EDT 1985
·
OSTI ID:5403490
Hybrid algorithm for solving sparse nonlinear systems of equations
Journal Article
·
Thu Dec 31 23:00:00 EST 1987
· Math. Comput.; (United States)
·
OSTI ID:5458148
Performance analysis of nonlinear equation solution techniques for chemical process design
Thesis/Dissertation
·
Thu Dec 31 23:00:00 EST 1987
·
OSTI ID:5172760