Estimation of the convergence points of a population using an individual pool
Jun Yu, Hideyuki Takagi · 2017
We employ an individual pool to increase the precision of the estimated convergence points of a population by using individual information from past generations. Better individuals from past generations are kept in the pool, and poorer individuals are replaced with new better individuals when the pool becomes full; convergence points for the population are thus estimated using those individuals from the pool that keeps excellent individuals in past generations. The estimated convergence points are used as elite individuals, and replace the worse individuals in current population to accelerate evolutionary computation. Besides the proposed basic pool storage mechanism, we further develop an extended version which enhances the interaction between an individual pool and the population in the latest generation. We evaluate these proposed methods using differential evolution and 14 benchmark functions. The experimental results show that introducing an individual pool can improve the convergence speed and accuracy with the same computational cost, and the extended version could further enhance the accelerated effect in almost all cases.