Enhancing AGDE Algorithm Using Population Size Reduction for Global Numerical Optimization

Citation:
Mohamed, A. K., A. W. Mohamed, E. Z. Elfeky, and M. Saleh, "Enhancing AGDE Algorithm Using Population Size Reduction for Global Numerical Optimization", The International Conference on Advanced Machine Learning Technologies and Applications (AMLTA2018), Cham, Springer International Publishing, pp. 62 - 72, 2018.

Date Presented:

2018

Abstract:

Adaptive guided differential evolution algorithm (AGDE) is a DE algorithm that utilizes the information of good and bad vectors in the population, it introduced a novel mutation rule in order to balance effectively the exploration and exploitation tradeoffs. It divided the population into three clusters (best, better and worst) with sizes 100p%, NP-2 * 100p% and 100p% respectively. Where p is the proportion of the partition with respect to the total number of individuals in the population (NP). AGDE selects three random individuals, one of each partition to implement the mutation process. Besides, a novel adaptation scheme was proposed in order to update the value of crossover rate without previous knowledge about the characteristics of the problems. This paper introduces enhanced AGDE (EAGDE) with non-linear population size reduction, which gradually decreases the population size according to a non-linear function. Moreover, a newly developed rule developed to determine the initial population size, that is related to the dimensionality of the problems.

Notes:

n/a

Tourism