An Efficient One Dimensional Search Procedure,
Abstract
Most efficient methods for unconstrained minimization utilize a one-dimensional search along directions generated by the method. If P is the function to be minimized, X the current vector of decision variables, and S the search direction, then the one-dimensional search problems is to choose alpha > 0 yielding the first local minimum of P(X + alphaS). The most popular one-dimensional search procedures for use in unconstrained minimization utilize quadratic or 2 point cubic interpolation of P. The paper describes a one-dimensional search based on quadratic and cubic interpolations. These are obtained using function values only. Computational results on interior and exterior penalty functions show that the method is considerably faster than any of the above mentioned techniques. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1973
- Accession Number
- AD0766506
Entities
People
- Arie Tamir
- Leon S. Lasdon
- Margery W. Ratner
- Richard L. Fox
Organizations
- Case Western Reserve University