NONCONVEX LINEAR PROGRAMMING.
Abstract
An algorithm for solving the 'nonconvex linear programming problem' Maximize z = cx, Subject to: (A sub i)x = or < (b sub i), 1 = 1,...,p for at least one i, x = or > 0 is given. The problem is viewed as one of n-dimensional geometry with the union of the p distinct solution sets a general nonconvex (possibly disconnected) polyhedral set. A continuous path in n-space is generated which passes from solution set to solution set as the optimum is sought. The core of the solution procedure is a 'linking linear program' which is a efficient, highly flexible substitute for Phase I of the simplex method. Given an arbitrary linear programming problem in n decision variables and having m constraints, the linking linear program solution approach can conceivably produce an optimum basic feasible solution with little more computational effort than minimum (n,m)+1 'simplex' pivot operations. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1970
- Accession Number
- AD0713484
Entities
People
- Robert E. Fricks
Organizations
- Case Western Reserve University