LP-form inclusion functions for global optimization (Q1174536): Difference between revisions

The paper presents a new approach to the bounding problem within branch- and-bound methods for global optimization where the objective and constraints are factorable \(C^ 1\) functions. For a given subregion generated by a branching strategy, a polyhedral envelope for each function is constructed, and the range of it is then computed by linear programs, determining bounds on the functions. ''Partial monotonicity'' is exploited to reduce the effective dimension of subregions.