IEEE Access, Journal Year: 2023, Volume and Issue: 11, P. 88711 - 88729

Published: Jan. 1, 2023

Differential Evolution(DE) is a widely used technique to tackle complex optimization problems owing its easy-implementation and excellent performance, nevertheless, the inborn weakness of crossover operation has not been solved even in recent state-of-the-art DE algorithms. There are two commonly schemes DE, exponential binomial crossover. The actually combination 1-point 2-point originated with GA, it positional bias because dependence on parameter separation. tackles by separating each dimension separately treating them independently, however, still exists from higher dimensional view, we name selection bias, that reason why QUATRE algorithm was proposed. evolution matrix primary component which solves previous variants suffer adaptation can be able escape some local optima optimization. Therefore, this paper proposes new better adaptations control parameter, moreover, perturbation mechanism firstly proposed for enhancement population diversity. main contributions our summarized as follows. First, generation proposed, obtain landscape objectives help jump out optima. Second, novel parameters also incorporating historical memory reduction. Third, enhance In order validate algorithm, intensive experiments conducted under 88 benchmark functions universal CEC2013, CEC2014, CEC2017 test suites comparison several variants, results support superiority.

Language: Английский