The nonlinear programming method of Wilson, Han, and Powell with an augmented Lagrangian type line search function. II. An efficient implementation with linear least squares subproblems (Q790581): Difference between revisions

[For part I see ibid. 38, 83-114 (1981; reviewed above).] This paper presents the results of numerical experiments for different versions of the method of R. B. Wilson, S.-P. Han and M. J. D. Powell. The following topics are studied and tested: the influence of the line search function on the numerical performance, the effect of a least square formulation of the quadratic subproblem, the dependence of the performance from different solution methods for the least squares subproblem, the usage of numerical differentiation, and the effect of an inverse update technique.