OMEGA one multi ethnic genetic approach

Abstract

The genetic algorithm (GA) is a quite efficient paradigm to solve several optimization problems. It is substantially a search technique that uses an ever-changing neighborhood structure related to a population which evolves according to a number of genetic operators. In the GA framework many techniques have been devised to escape from a local optimum when the algorithm fails in locating the global one. To this aim we present a variant of the GA which we call OMEGA (One Multi Ethnic Genetic Approach). The main difference is that, starting from an initial population, k different sub-populations are produced at each iteration and they independently evolve in k different environments. The resulting sub–populations are then recombined and the process is iterated. Our basic algorithmic scheme is tested on a recent and well-studied variant of the classic problem of the minimum spanning tree: the Minimum Labeling Spanning Tree problem. We compare our algorithm with several approaches drawn from the literature. The results are encouraging in view of future application of OMEGA to other classes of problems

Heuristics for the strong generalized minimum label spanning tree problem

“…In this section, we compare the performances of the three proposed heuristic approaches on a set of benchmark instances. In order to generate test cases, we started from a dataset used in previous works for the base MLST problem . The instances that we considered are randomly generated graphs having a number of nodes ∣ V ∣ equal to 20, 30, 40, 50, 100, or 200, a number of labels ∣ L ∣ equal to the number of nodes, and a number of edges ∣ E ∣ equal to , where d is a density measure equal to 0.2, 0.5, or 0.8.…”

“…The authors also showed this new bound to be tight. Metaheuristics for the problem were proposed in several works . In , the authors presented a single‐commodity flow mathematical model, and showed that the formulation obtained by relaxing integrality on the arc selection variables always provides the optimal solution value for the original unrelaxed problem.…”

A multiethnic genetic approach for the minimum conflict weighted spanning tree problem

“…The Mega is a technique developed for the GAs which is aimed at reducing the probability of remaining trapped at a local minimum [4]. The main idea behind the algorithm is to split a starting population in k different subpopulations that, independently, evolve in k different environments.…”

“…Actually, the idea of allowing the simultaneously evolve of k different populations was already adopted in parallel GAs as the Island Model (see [12,19]). However, in Mega each population is characterized by its own fitness function that is appropriately selected to diversify the evolution and to carry out a better exploration of the solution space (for more details see [4]).…”

Preface to the Special Issue “Recent developments in non-linear and global optimization”

“…• The paper [6] written by Cerrone, Cerulli, and Gaudioso presents a genetic algorithm called by the authors OMEGA (One Multi Ethnic Genetic Approach). The main difference with respect to existing genetic methods is that, starting from an initial population, k different sub-populations are produced at each iteration and they independently evolve in k different environments.…”