New approach for solving large residual nonlinear least squares problems
This paper is concerned with the solution of the nonlinear least squares problem. A continuation method is used to develop a new framework for the model trust region approach for solving nonlinear least squares problems. This framework gives a motivation for the direct selection of the trust region parameter. It also provides a natural safeguard for trust region methods which leads to a very robust algorithm. A class of algorithms based on the continuation method will be presented. In addition, the implementation details for one member of the new class will be examined. The convergence and descent properties of this algorithm will also be discussed. Numerical evidence will be given which shows that the new algorithms are competitive with existing model trust region algorithms. For large residual problems or problems in which a good initial starting guess is not available, the performance of the new algorithm is very promising. 24 references, 3 tables.
- Research Organization:
- Sandia National Labs., Albuquerque, NM (USA)
- DOE Contract Number:
- AC04-76DP00789
- OSTI ID:
- 6211038
- Report Number(s):
- SAND-84-1564; ON: DE85002555
- Country of Publication:
- United States
- Language:
- English
Similar Records
Software performance on nonlinear least-squares problems
New algorithm for constrained nonlinear least-squares problems. Part I