Adaptive differential evolution with fitness-based crossover rate for global numerical optimization

Abstract Differential evolution (DE) is one of the most efficient algorithms (ES) for dealing with nonlinear, complicated and difficult global optimization problems. The main contribution this paper can be summarized in three directions: Firstly, a novel crossover rate (CR) generation scheme based on zscore value fitness, named fcr , introduced. For minimization problem, proposed CR strategy always assigns smaller to individual fitness value. Therefore, parameters individuals better are inherited by their offspring high probability. In second direction, control adjusted unused bimodal settings which each parameter setting selected according status individual. third direction our work introducing L1 norm distance as weights updating mean scale factor. Theoretically, compared L2 norm, L1-norm more suppress outliers difference vector. These modifications first integrated mutation JADE, then modified version, JADEfcr, proposed. addition, improve ability further, another variant LJADEfcr using linear population reduction mechanism considered. So confirm examine performance JADEfcr LJADEfcr, numerical experiments conducted 29 problems defined CEC2017 benchmark. its experimental results made comparison twelve state-of-the-art algorithms. comparative study demonstrates that terms robustness, stability solution quality, highly competitive these well-known other nine powerful including four recent five top competition. Experimental indicate superior statistically excellent quality obtained solutions.