On the asymptotic convergence of differential evolution in continuous spaces: a control theoretic approach

Theoretical analysis of the properties of evolutionary algorithms is very important to understand their search behaviors and to develop more efficient algorithms. This article investigates the convergence properties of a canonical Differential Evolution (DE) algorithm with DE/rand/1 type mutation and binomial crossover. For simplicity and to provide an insight into the heuristics of the algorithm, the analysis has been done by assuming a single-dimensional fitness function f(x) . The analysis is independent of the nature of the objective function as long as it remains real-valued and possesses an unique global optimum (it may have multiple local optima as well).