Abstract
We present a generalized Newton method and a quasi-Newton method for solving H(x):=F(? c (x))+x-? c (x)=0, when C is a polyhedral set. For both the Newton and quasi-Newton methods considered here, the subproblem to be solved is a linear system of equations per iteration. The other characteristics of the quasi-Newton method include: (i) a Q-superlinear convergence theorem is established without assuming the existence of H' at a solution x* of H(x)=0; (ii) only one approximate matrix is needed; (iii) the linear independence condition is not assumed; (iv) Q-superlinear convergence is established on the original variable x; and (v) from the QR-factorization of the kth iterative matrix, we need at most O((1+2|J k |+2|L k |)n 2 ) arithmetic operations to get the QR-factorization of the (k+1)th iterative matrix.
Original language | English |
---|---|
Pages (from-to) | 659-676 |
Number of pages | 18 |
Journal | Journal of Optimization Theory and Applications |
Volume | 94 |
Issue number | 3 |
DOIs | |
Publication status | Published - 1 Sept 1997 |
Externally published | Yes |
Keywords
- Newton methods
- Normal maps
- Q-superlinear convergence
- quasi-Newton methods
ASJC Scopus subject areas
- Control and Optimization
- Management Science and Operations Research
- Applied Mathematics