Differential evolution enhanced with multiobjective sorting-based mutation operators.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Evolutionary Computation

Background:

• Differential evolution (DE) is a population-based evolutionary algorithm known for its simplicity and power.
• Current DE mutation operators often lack selective pressure and ignore population diversity.
• Random parent selection in DE can lead to suboptimal exploration and exploitation.

Purpose of the Study:

• To develop a DE framework that effectively utilizes both fitness and diversity information.
• To introduce a multiobjective sorting-based mutation operator for improved parent selection.
• To enhance the performance of DE algorithms by balancing exploration and exploitation.

Main Methods:

• A novel mutation operator employing nondominated sorting for fitness and diversity assessment.
• Proportional parent selection based on rankings derived from fitness and diversity.
• Integration of the proposed operator into original and advanced DE variants.

Main Results:

• The proposed operator significantly enhances the performance of most DE algorithms studied.
• Experimental results on 48 benchmark functions and 12 real-world problems validate the approach.
• Simultaneous consideration of fitness and diversity leads to a better balance between exploration and exploitation.

Conclusions:

• The multiobjective sorting-based mutation operator is an effective enhancement for DE algorithms.
• Exploiting both fitness and diversity information improves evolutionary computation performance.
• The proposed method offers a robust strategy for optimizing complex problems using DE.