LCLSQ: an implementation of an algorithm for linearly constrained linear least-squares problems. [For IBM 370 and 3033, in FORTRAN]
This report describes the implementation of an algorithm of Stoer and Schittkowski for solving linearly constrained linear least-squares problems. These problems arise in many areas, particularly in data fitting where a model is provided and parameters in the model are selected to be a best least-squares fit to known experimental observations. By adding constraints to the least-squares fit, one can force user-specified properties on the parameters selected. The algorithm used applies a numerically stable implementation of the Gram-Schmidt orthogonalization procedure to deal with a factorization approach for solving the constrained least-squares problem. The software developed allows for either a user-supplied feasible starting point or the automatic generation of a feasible starting point, redecomposition after solving the problem to improve numerical accuracy, and diagnostic printout to follow the computations in the algorithm. In addition to a description of the actual method used to solve the problem, a description of the software structure and the user interfaces is provided, along with a numerical example. 3 figures, 1 table.
- Research Organization:
- Argonne National Lab. (ANL), Argonne, IL (United States)
- DOE Contract Number:
- W-31-109-ENG-38
- OSTI ID:
- 6778440
- Report Number(s):
- ANL-80-116
- Country of Publication:
- United States
- Language:
- English
Similar Records
FC; LSEI; WNNLS; constrained least squares fit. [CDC6600,7600,CYBER175 (designed to be machine-independent); FORTRAN IV]
Fast Combinatorial Algorithm for the Solution of Linearly Constrained Least Squares Problems