Fast QR decomposition for weighted least squares problems

Conference · Fri Dec 31 04:00:00 EST 1993

OSTI ID:54433

Anda, A A; Park, H ^[1]

---

1. Univ. of Minnesota, Minneapolis, MN (United States)

We present algorithms which apply dynamically scaled fast plane rotations to the QR decomposition for stiff least squares problems. When an equality constrained linear least square problem is solved via extreme weighting of the constraint equations, a very stiff matrix is generated. Numerical test results show that the accuracy of our algorithm compares favorably with that of the Givens rotation based algorithm while the Householder method may produce very sensitive results. Moreover, both fast and standard Givens rotation based algorithms produce very accurate results regardless of row sorting and even with extremely large weights.

• OSTI ID: 54433
• Report Number(s): DOE/ER/25151--1-Vol.1; CONF-930331--Vol.1; CNN: Contract DAAL02-89-C-0038; Grant CCR-9209726
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
ALGORITHMS
EFFICIENCY
MATRICES
NUMERICAL SOLUTION
PARALLEL PROCESSING
VARIATIONAL METHODS