We propose to further develop parallel approaches to solving mixed integer and stochastic programs that build on the computationally successful basic ideas (heuristic lattice basis reduction; and the use of random walk points) introduced in our previous research. Specifically, we intend to test heuristics (i) for using approximate adjoint lattices when generating branching hyperplanes; (ii) develop cut generation methods that use information from multiple points and disjunctions for generating improved quality cuts; (iii) develop branch-and-cut algorithms for MINLP when the constraint functions are not differentiable, particularly from the viewpoint of stochastic mixed integer programming.
|Effective start/end date||6/1/15 → 8/31/16|
- Office of Naval Research (N00014-15-1-2226)
Mixed integer programming