Truncated-newtono algorithms for large-scale unconstrained optimization
- 505 Downloads
We present an algorithm for large-scale unconstrained optimization based onNewton's method. In large-scale optimization, solving the Newton equations at each iteration can be expensive and may not be justified when far from a solution. Instead, an inaccurate solution to the Newton equations is computed using a conjugate gradient method. The resulting algorithm is shown to have strong convergence properties and has the unusual feature that the asymptotic convergence rate is a user specified parameter which can be set to anything between linear and quadratic convergence. Some numerical results on a 916 vriable test problem are given. Finally, we contrast the computational behavior of our algorithm with Newton's method and that of a nonlinear conjugate gradient algorithm.
Key wordsUnconstrained Optimization Modified Newton Methods Conjugate Gradient Algorithms
Unable to display preview. Download preview PDF.
- O. Axelsson, “Solution of linear systems of equations: Iterative methods”, in: V.A. Barker, ed.,Sparse matrix techniques. (Springer-Verlag, New York, 1976) pp. 1–51.Google Scholar
- R. Chandra, “Conjugate gradient methods for partial differential equations”, Ph.D. dissertation, Yale University (New Haven, CT, 1978), Also available as Department of Computer Science Research Report No. 129.Google Scholar
- R. Chandra, S.C. Eisenstat and M.H. Schultz, “The modified conjugate residual method for partial differential equations”, in: R. Vichnevetsky, ed.,Advance in computer methods for partial differential equations II, Proceedings of the Second International Symposium on Computer Methods for Partial Differential Equations, Lehigh University, Bethlehem, PA (International Association for Mathematics and Computers in Simulation, June 1977) pp. 13–19.Google Scholar
- R. S. Dembo and T. Steihaug, “A test problem for large-scale unconstrained minimization”, School of Organization and Management, Yale University (New Haven, CT) Working Paper Series B (in preparation).Google Scholar
- N.K. Garg and R.A. Tapia, “QDN: A variable storage algorithm for unconstrained optimization”, Technical Report, Department of Mathematical Sciences, Rice University (Houston, TX, 1977).Google Scholar
- P.E. Gill and W. Murray, “Safeguarded steplength algorithm for optimization using descent methods”, Technical Report NPL NA 37, National Physical Laboratory (1974).Google Scholar
- P.E. Gill, W. Murray and S.G. Nash, “A conjugate-gradient approach to Newton-type methods”, presented at ORSA/TIMS Joint National Meeting (Colorado Springs, November 1980).Google Scholar
- T. Steihaug, “Quasi-Newton methods for large scale nonlinear problems”, Ph.D. dissertation, Yale University (New Haven, CT, 1980). Also available as Working Paper, Series B No. 49.Google Scholar