Theorems related to symmetric nonlinear decomposition

Decomposition methods can provide the rescue from the “curse of dimensionality”, which often prevents the successful numerical solution of large scale nonlinear mathematical programming problems. A symmetric nonlinear decomposition theory has been elaborated by T.O.M. Kronsjö (4) as an extension of a theory by the same author (3). The stringent proof of the convergence of this decomposition algorithm requires some results on necessary optimality conditions for certain mathematical programming problems. In this paper we state and prove some theorems providing these results.     

Theorems on the decomposition of a large nonlinear convex separable economic system in the dual direction

Large mathematical programming problems often arise as the result of the economic planning process. When such a problem is not only large, but nonlinear as well, there is a need to make it more manageable by breaking it down into several smaller and more easily handled subproblems. The subproblems are solved separately with the coordination activity carried out by a “master problem”. A decomposition method can be seen as a “dialogue” between the master problem and the subproblems, where the flow of information back and forth between the former and the latters results in a series of approximations converging to the solution of the overall problem. Such a decomposition method was elaborated by Benders [1] for linear programmes and generalized to nonlinear convex separable programmes by Kronsjö [4] and by Geoffrion [3]. After considering our basic nonlinear programming problem from a two-stage minimization point of view, we review the Kronsjö nonlinear decomposition algorithm. Then we establish some properties of a function related to this algorithm.