Scaling symmetric rank one update for unconstrained optimization
A basic disadvantage to the symmetric rank one (SR1) update is that the SR1 update may not preserve positive definiteness when starting with a positive definite approximation. A simple remedy to this problem is to restart the update with the initial approximation mostly the identity matrix whenever...
Saved in:
Main Authors: | , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Universiti Utara Malaysia
2003
|
Subjects: | |
Online Access: | http://repo.uum.edu.my/369/1/Malik_Abu_Hassan.pdf http://repo.uum.edu.my/369/ http://ijms.uum.edu.my |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | A basic disadvantage to the symmetric rank one (SR1) update is that the SR1 update may not preserve positive definiteness when starting with a positive definite approximation. A simple remedy to this problem is to restart the update with the initial approximation mostly the identity matrix whenever this difficulty arises. However, numerical experience shows that restart with the identity matrix is not a good choice. Instead of using the identity matrix we used a positive multiple of the identity matrix. They Used positive scaling factor is the optimal solution of the measure defined by the problem - maximize the determinant subject to a bound of 1 on the largest eigenn value. This measure is motivated by considering the volume of the symmetric diference of the two ellipsoids, which arise from the current and updated quadratic models in quasi-Newton methods. A replacement in the form of positive multiple of identity matrix is provided for the SRI when it is not positive definite. Our experiments indicate that with such simple scale, the efectiveness of the SRI method is increased dramatically. |
---|