The nonlinear programming method of Wilson, Han, and Powell with an augmented Lagrangian type line search function. I. Convergence analysis (Q790580): Difference between revisions

Comparative studies of codes for solving continuously differentiable nonlinear programs indicate that quadratic approximation methods are efficient and reliable. In this paper the author presents a convergence analysis of the method of \textit{R. B. Wilson, S.-P. Han} and \textit{M. J. D. Powell} [cf. the author's book: Nonlinear programming codes. Information, tests, performance (1980; Zbl 0435.90063)]. This analysis results in overcoming some of the theoretical disadvantages and in improvement of numerical performance of the method. This is achieved by replacing the exact \(L_ 1\)-penalty function by a differentiable augmented Lagrange function for the line search computation.