แนวคิดหลัก
A novel reconstructed differential evolution (RDE) algorithm that combines effective strategies from recent advanced DE variants to efficiently solve single-objective bound constrained optimization problems.
บทคัดย่อ
The paper proposes a new differential evolution (DE) variant called Reconstructed Differential Evolution (RDE) to solve single-objective bound constrained optimization problems. RDE combines several effective strategies from recent advanced DE algorithms:
External archive to store historical solutions
DE/current-to-order-pbest/1 mutation strategy that sorts and recombines differential terms
Adaptive hybridization of DE/current-to-order-pbest/1 and DE/current-to-pbest/1 mutation strategies
Extended rank-based selective pressure strategy for mutation term selection
Success-history based parameter adaptation for scale factor F and crossover rate Cr
Linear population size reduction
Cauchy perturbation to enhance population diversity
The authors tested RDE on the CEC2024 benchmark suite and compared it against several state-of-the-art DE variants. The experimental results show that RDE outperforms the competitors on the majority of the test functions, demonstrating its superior performance in solving complex single-objective bound constrained optimization problems.
สถิติ
The mean and standard deviation of the function values obtained by RDE and the competitor algorithms on the 29 CEC2024 benchmark functions are provided.
คำพูด
"RDE demonstrates superiority over competitive algorithms on CEC2024."
"Compared with LSHADE-RSP, iLSHADE-RSP, HSES, EBOwithCMAR and LSHADE, RDE achieves better solutions on 13, 15, 11, 14 and 17 problems respectively."