Metaheuristic approaches based on neighborhood search escape local optimality by applying predefined rules and constraints, such as tabu restrictions (in tabu search), acceptance criteria (in simulated annealing), and shaking (in VNS). We propose a general approach that attempts to learn...