The Reactive Tabu Search

Abstract

We propose an algorithm for combinatorial optimization where an explicit check for the repetition of configurations is added to the basic scheme of Tabu search. In our Tabu scheme the appropriate size of the list is learned in an automated way by reacting to the occurrence of cycles. In addition, if the search appears to be repeating an excessive number of solutions excessively often, then the search is diversified by making a number of random moves proportional to a moving average of the cycle length. The reactive scheme is compared to a “strict” Tabu scheme that forbids the repetition of configurations and to schemes with a fixed or randomly varying list size. From the implementation point of view we show that the Hashing or Digital Tree techniques can be used in order to search for repetitions in a time that is approximately constant. We present the results obtained for a series of computational tests on a benchmark function, on the 0-1 Knapsack Problem, and on the Quadratic Assignment Problem. INFORMS Journal on Computing, ISSN 1091-9856, was published as ORSA Journal on Computing from 1989 to 1995 under ISSN 0899-1499.

Out-of-the-Box and Custom Implementation of Metaheuristics. A Case Study: The Vehicle Routing Problem with Stochastic Demand

“…This approach is particularly appealing when one is supposed to solve one single instance, typically large and complex. One of the most influential descriptions of on-line adjustment of the parameters of an algorithm has been given by Battiti and Tecchiolli [60]. The authors introduced a tabu search where the length of the tabu list is optimized on-line.…”

Analysis of reactive search optimisation techniques for the maximum clique problem and applications

“…Reactive Search [14,9] advocates the use of machine learning to automate the parameter tuning process and make it an integral and fully documented part of the algorithm. Learning is performed on-line, and therefore task-dependent and local properties of the configuration space can be used.…”

“…CLS pseudocode is depicted in Listing 6.1. It alternates between an iterative improvement phase (lines 7-11) and a plateau search phase (lines [12][13][14]. In the former a maximal solution is constructed by repeatedly expanding the current configuration C by selecting a candidate vertex from PossibleAdd.…”

“…Finally, a Restart mechanism is provided (lines [14][15]: if the best solution is not improved in a while, then the algorithm is restarted, so that Listing 7.2: ChooseNode chooses the non-prohibited node having the best chance to lead to better cliques in the future; if no nodes can be added, it picks one for removal. function ChooseNode(L,R, P, T, t, E, F E )…”

Heuristic algorithms and learning techniques: applications to the graph coloring problem

“…The term "reactive mechanisms" refers to using an on-line "feedback" loop for tuning parameters according to the current state of the search. Reactive Tabu Search [Battiti and Tecchiolli, 1994] is a representative example of such concepts; however, algorithms can also react on the neighborhood, on the annealing schedule, or on the objective function-see Chapters 2-5 in . This subdirection is closely intertwined with the others because a parameter change can trigger a complete transformation of the algorithm.…”