Mathematical Programming Solver Based On Local Search

This book covers local search for combinatorial optimization andits extension to mixed-variable optimization. Although not yetunderstood from the theoretical point of view, local search is theparadigm of choice for tackling large-scale real-life optimizationproblems. Today's end-users demand interactivity withdecision support systems. For optimization software, this meansobtaining good-quality solutions quickly. Fast iterativeimprovement methods, like local search, are suited to satisfyingsuch needs. Here the authors show local search in a new light, inparticular presenting a new kind of mathematical programmingsolver, namely LocalSolver, based on neighborhood search.First, an iconoclast methodology is presented to design andengineer local search algorithms. The authors' concernregarding industrializing local search approaches is of particularinterest for practitioners. This methodology is applied to solvetwo industrial problems with high economic stakes. Software basedon local search induces extra costs in development and maintenancein comparison with the direct use of mixed-integer linearprogramming solvers. The authors then move on to present theLocalSolver project whose goal is to offer the power of localsearch through a model-and-run solver for large-scale 0-1 nonlinearprogramming. They conclude by presenting their ongoing and futurework on LocalSolver toward a full mathematical programming solverbased on local search.