Interval Differential Evolution Using Structural Information of Global Optimization Problems

Differential Evolution (DE) algorithms are a promising strategy to Numerical Constrained Global Optimization Problems (NCOP). Most DE recent variants are applied to black-box optimization problems, where the analytical structure of the NCOP instance is unknown. In this paper we present an Interval Differential Evolution (InDE) algorithm that explores the structural information of the problem. The instance structure is represented by a hypergraph Epiphytic decomposition, where the variables are intervals. InDE algorithm is based on several strategies used in state-of-the-art DE implementations. Based on structural information, our approach extracts a subset of variables of the instance that are critical to the search process. The DE population individuals encode only this subset of variables. The other variables of the instance are valuated by a linear cost constraint propagation over the hypergraph structure. Our experiments show that the use of structural information associated with interval local consistency techniques significantly improves the performance of DE algorithm.