Mats Werme, KTH
E-mail: werme@math.kth.se
Thereafter, two different methods for efficient sensitivity calculations are presented. These methods are applied in the context of neighbourhood search methods where one repeatedly jumps to the "best" neighbour of the current design until a local optimum has been found. Next, the neighbourhood search methods are generalized and the nonlinear integer programming problems are solved to provable local optimality by the use of sequential integer programming methods. The key idea in the sequential integer programming methods is to generate and solve a sequence of "easier" subproblems where the sequence of solutions will converge to a local optimum of the original problem.
Both theoretical and numerical results will be presented.