FD-DE: Differential Evolution with fitness deviation based adaptation in parameter control.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Evolutionary Computation

Background:

• Differential Evolution (DE) is a powerful stochastic optimization method.
• Existing DE variants possess notable limitations.
• Advanced optimization techniques are crucial for complex problems.

Purpose of the Study:

• To propose a novel and powerful DE variant for single-objective numerical optimization.
• To address the weaknesses of current state-of-the-art DE algorithms.
• To enhance the performance and robustness of DE.

Main Methods:

• An enhanced wavelet basis function for scale factor generation.
• A hybrid trial vector generation strategy incorporating perturbation and t-distribution.
• Fitness deviation-based parameter control and a novel diversity indicator with a restart scheme.

Main Results:

• The proposed DE variant achieved significant improvements on a large test suite (130 benchmarks).
• Superior performance was demonstrated compared to several well-known state-of-the-art DE variants.
• Validation on real-world optimization applications confirmed the algorithm's superiority.

Conclusions:

• The novel DE variant offers a substantial advancement in single-objective numerical optimization.
• The proposed enhancements effectively address limitations of existing DE algorithms.
• The algorithm shows strong potential for practical application in complex optimization tasks.