Delayed improvement local search

Abstract Local search is a fundamental tool in the development of heuristic algorithms. A neighborhood operator takes current solution and returns set similar solutions, denoted as neighbors. In best improvement local search, neighboring solutions replaces each iteration. On other hand, first only explored until any improving found, which then solution. this work we propose new strategy for that attempts to avoid low-quality optima by selecting iteration neighbor has fewest possible attributes common with optima. To end, it uses inequalities previously used optimality cuts context integer linear programming. The novel method, referred delayed implemented evaluated using travelling salesman problem 2-opt max-cut 1-flip test cases. Computational results show strategy, while slower, obtains better compared traditional strategies. comparison favourable experiments fixed computation time or target.