IM-NSGAII: A novel approach to boost convergence speed and population diversity in multi-objective optimization

Area of Science:

• Multi-objective evolutionary algorithms
• Computational intelligence
• Optimization techniques

Background:

• Convergence speed and population diversity are critical in multi-objective evolutionary algorithms (MOEAs).
• The standard NSGAII algorithm struggles with maintaining population diversity on complex Pareto fronts.
• Addressing these limitations is crucial for advancing MOEA performance.

Purpose of the Study:

• To propose an improved NSGAII algorithm (IM-NSGAII) that enhances both convergence speed and population diversity.
• To overcome the limitations of the standard NSGAII in handling complex Pareto fronts.
• To provide a more robust and efficient multi-objective optimization approach.

Main Methods:

• Incorporation of a population evaluation technique post non-dominated sorting for parent selection.
• Implementation of a sparse population strategy with a high-pressure criterion for local exploration and diversity enhancement.
• Introduction of a difference operator to facilitate information exchange among sparse individuals, improving convergence.

Main Results:

• The IM-NSGAII algorithm demonstrated significant improvements in population diversity compared to existing algorithms.
• Experimental results confirmed enhanced convergence speed of IM-NSGAII on benchmark problems.
• IM-NSGAII outperformed five widely used algorithms on ZDT, DTLZ, MaF, and WFG datasets.

Conclusions:

• The proposed IM-NSGAII effectively addresses the diversity maintenance and convergence speed issues of the standard NSGAII.
• IM-NSGAII offers a promising advancement for multi-objective evolutionary optimization, especially for complex problems.
• The novel strategies employed in IM-NSGAII contribute to superior performance in both diversity and convergence.