Stability Analysis of DESA Optimization Algorithm

摘要

This paper investigates the dynamics of the hybrid evolutionary optimization algorithm, Differential Evolution-Simulated Annealing (DESA) algorithm with the binomial crossover and SA-like selection operators. A detailed mathematical framework of the operators of the DESA/rand/1/bin algorithm is provided to characterize the behavior of the DESA-population system. In DESA, the SA-like selection operation provides a nonzero probability of accepting a deteriorated solution that decreases with a sufficient number of generations. This paper shows that the system defined by the DESA-population is stable. Moreover, the DESA-population system time constant, learning and momentum rates are dependent on the value of the crossover constant and the probability of accepting deterioration in the quality of the objective function.