JDF-DE: a differential evolution with Jrand number decreasing mechanism and feedback guide technique for global numerical optimization

Differential Evolution (DE) is a readily comprehensible and highly powerful intelligent optimized method for numerical optimization. The performance of DE significantly depends on its parameters and strategies generating both mutation vector and trial vector. To further enhance its exhibition, we propose a new DE variant called JDF-DE based on JADE by introducing the improved parameter approach with weight and crossover strategy with Jrand number decreasing mechanism and feedback guide technique. The new way for updating parameter μ_CR and μ_F brings fitness value to generate more reasonable parameters with the fixed orientation during evolution. Meanwhile, Levy distribution is used to complete the adaptive distribution of CR when the population has a high clustering intensity so that solutions escape from the local optimal value. Jrand number decreasing mechanism is embedded to crossover operation to strengthen population diversity instead the number of Jrand equals 1 in primary DE algorithm. Feedback guide method is utilized to determine the step size for Jrand number to advance the ability that JDF-DE searches the optimum value. In order to investigate performance of JDF-DE, In order to analyze performance of JDF-DE, 29 benchmark functions from CEC2017 on real parameter optimization are employed to verify the validity of JDF-DE for solving complex high-dimensional problems. The experimental results show that JDF-DE is better than, or at least comparable with several state-of-the-art DE variants including DE variants JADE, SinDE, TSDE, AGDE, and EFADE and non-DE variants TSA, SHO, GWO, MVO, SCA, and GSA in the global numerical optimization problems.