An improved parent-centric mutation with normalized neighborhoods for inducing niching behavior in differential...

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Evolutionary Computation

Background:

• Real-world problems often require finding multiple optimal solutions.
• Evolutionary multimodal optimization algorithms are designed for this purpose.
• Differential evolution (DE) is effective for single-solution optimization.

Purpose of the Study:

• To develop a stable and efficient niching scheme for differential evolution (DE).
• To address the challenge of problem-dependent parameters in niching.
• To enhance the capability of DE for multimodal optimization.

Main Methods:

• Integration of a parent-centric mutation operator with a synchronous crowding replacement rule into DE.
• Development of a parameter-free niching approach.
• Utilizing local neighborhoods to maintain population diversity.

Main Results:

• The proposed niching DE algorithm demonstrates stable and efficient niching behavior.
• The algorithm avoids the need for problem-dependent control parameters like niche radius.
• Consistent statistical superiority was observed compared to 13 state-of-the-art niching evolutionary algorithms.

Conclusions:

• The proposed niching scheme significantly improves DE's ability to find multiple optimal solutions.
• The parameter-free design enhances the algorithm's practical applicability.
• The novel approach offers a robust and superior alternative for multimodal optimization problems.