Successive Search Methods for a Canonical DC Programming Problem

摘要

In this talk, we propose successive search methods for solving a canonical dc programming problem in the case of n ≥ 3. The algorithms generate a sequence of finite sets contained in the boundary of a polytope on the orthogonal complement space of the vector defining the objective function. Moreover, at each iteration, the algorithms search a boundary point of the feasible set on the two-dimensional subspace spanned by a feasible solution and the midpoint between two elements selected from the finite set. It is shown that the algorithms have the global convergence property.