Numerical Aspects of Trajectory Algorithms for Nonlinearly Constrained Optimization.
Abstract
This paper discusses two algorithms for nonlinearly constrained optimization. These algorithms -- the penalty and barrier trajectory algorithms -- are based on an examination of the trajectories of approach to the solution that characterize the quadratic penalty function and the logarithmic barrier function, respectively. Although closely related in principle, the two algorithms display important differences in their implementation as well as in the properties of the generated iterates. The discussion will emphasize the numerical aspects of implementation of the trajectory algorithms, with particular attention to the choice of reliable methods for carrying out the required computations. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 1976
- Accession Number
- ADA044903
Entities
People
- Margaret H. Wright
- Walter Murray
Organizations
- Stanford University