Domain-independent local search for linear integer optimization
Joachim P. Walser · 1998
Integer and combinatorial optimization problems constitute a major challenge for algorithmics. They arise when a large number of discrete organizational decisions have to be made, subject to constraints and optimization criteria. This thesis describes and investigates new domain-independent local search strategies for linear integer optimization. We introduce WSAT(OIP), an integer local search method which operates on an algebraic problem representation. WSAT(OIP) generalizes Walksat, a successful local search procedure for propositional satisfiability (SAT), to more expressive constraint systems. For this purpose, we introduce over-constrained integer programs (OIPs), a constraint class which is closely related to integer programs. OIP allows for a natural generalization of the principles of SAT local search to integer optimization. Further, it will be shown that OIPs are a special case of integer linear programs and permit combinations with linear programming for bound computation, initialization by rounding, search space reduction, and feasibility testing. The representation