Article ID: | iaor20033316 |
Country: | Japan |
Volume: | 46 |
Issue: | 1 |
Start Page Number: | 1 |
End Page Number: | 21 |
Publication Date: | Mar 2003 |
Journal: | Journal of the Operations Research Society of Japan |
Authors: | Yabe Hiroshi, Ogasawara Hideho |
Keywords: | optimization |
We are concerned with nonlinear least squares problems. It is known that structured quasi-Newton methods perform well for solving these problems. In this strategy, two kinds of factorized structured quasi-Newton methods have been independently proposed by Yabe and Takahashi, and Sheng and Zou. Sheng and Zou introduced a Broyden–Fletcher–Goldfarb–Shanno (BFGS)-like update by considering how the normal equation based on an affine model may consist with the Newton equation, and dealt with a hybrid method that combines the Gauss–Newton method and their BFGS-like method. In this paper, we deal with the Sheng–Zou–Broyden family proposed by Yabe, which is an extension of the update of Sheng and Zou to the Broyden-like family. Local and