Hybrid Metaheuristics for the Far From Most String Problem

A GRASP-based memetic algorithm with path relinking for the far from most string problem

“…For GRASP FFR and GRASP Mou algorithms, mean improvement percentage of solutions (imp.%) obtained by MA GRASP þ HC þ PR with respect to the ones obtained by these algorithms is also shown. Mousavi et al (2012) and Ferone et al (2013) is also favorable to the MA, which provides average improvements of 8% and 74%, respectively on the hardest random problem instances and of 8% and 60% on the hardest real-world instances. Statistical tests confirm the significance of these improvements.…”

“…This GA provided average improvements of 54-257% over the previous GRASP, albeit at a higher computational cost. Finally, Ferone et al (2013) proposed a variable neighborhood search (VNS) metaheuristic and its use both as a stand-alone algorithm or within GRASP (as a local-search component). Also, path relinking is used as an intensification procedure, either in GRASP, VNS or hybrids thereof.…”

“…Initialization of individuals in the population was done by means of a Greedy Randomized Adaptive Search Procedure (GRASP) similar to the one proposed in Ferone et al (2013). Pseudo-code for this procedure is shown in Fig.…”

“…This symbol is chosen uniformly at random from a restricted candidate list (RCL) that includes candidate symbols. A parameter α (0 r αr1) is introduced as a generalization of the original GRASP component of the algorithm by Ferone et al (2013), in order to control the degree of greediness used to generate the solution: smaller values for this parameter imply the generation of greedier solutions and larger values introduce more randomness. For instance, if α is 0, so is β, and then only the least frequent symbol is included in the RCL (this would correspond to a greedy selection strategy where the least frequent symbol for each position would be selected).…”

Hybridizations of GRASP with path relinking for the far from most string problem

“…Ferone et al. () designed the following pure and hybrid multistart iterative heuristics: a pure GRASP, inspired by Festa (), a GRASP that uses forward path relinking for intensification, a pure variable neighborhood search (VNS), a VNS that uses forward path relinking for intensification, a GRASP that uses VNS to implement the local search phase, and a GRASP that uses VNS to implement the local search phase and forward path relinking for intensification. …”