Stability and Convergence in Mathematical Programming.
Abstract
The authors consider the point-to-set map (S sub b) = (x / g(x) < or = b), where g(.) is a function from (E sup n) to (E sup m) and b is an m-vector. Previous work by Evans and Gould, Greenberg and Pierskalla, and others has established necessary and sufficient conditions for this map to be upper semi-continuous or lower semi-continuous. Here the authors consider the stronger notion of linear continuity, which requires that changes in (S sub b) as a function of b satisfy a Lipshitz condition. Necessary conditions are presented for which (S sub b) has this property. The authors then consider the class of problems of maximizing a function over (S sub b) and give conditions under which the optimal values of the objective function satisfies a Lipshitz condition in terms of b. Conditions are also given for the sets of epsilon-optimal solutions to be linearly continuous in b. Both convex and nonconvex problems are considered. Some extensions are made to functional perturbations and other areas. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1973
- Accession Number
- AD0767978
Entities
People
- Donald M. Topkis
- Michael H. Stern
Organizations
- University of California, Berkeley