A Novel Approach Of Applying The Differential Evolution To Spatial Discrete Data

The Differential Evolution (DE) is a powerful bioinspired algorithm searching optimal solutions. The actual DE modifications can handle the real, integer and discrete valued problems. The values of the discrete-valued variables represent the integer indices addressing the discrete samples in the ordered array. The optimization in unordered samples leads to a random search. This paper proposes a novel modification dealing with d-dimensional discrete vertices. A vertex hashing is used to strengthen the local properties of a dataset and to improve the spatial convergence of the evolution.