A Collaborative Neurodynamic Approach to Multiobjective Optimization.

Area of Science:

• Computational Intelligence
• Optimization Theory
• Artificial Neural Networks

Background:

• Multiobjective optimization aims for Pareto-optimal solutions and uniform distribution along the efficient frontier.
• Existing algorithms often struggle to balance solution optimality and diversity effectively.

Purpose of the Study:

• To present a collaborative neurodynamic approach for multiobjective optimization.
• To achieve both Pareto optimality and solution diversity simultaneously.
• To enhance the characterization of the efficient frontier.

Main Methods:

• Scalarization of multiple objectives using a weighted Chebyshev function.
• Utilization of multiple projection neural networks for Pareto-optimal solution search.
• Integration of particle swarm optimization (PSO) for reinitialization and hypervolume (HV) maximization.
• A holistic approach to diversify solutions by maximizing hypervolume.

Main Results:

• The proposed neurodynamic approach demonstrated superior performance in terms of hypervolume (HV) and inverted generational distance.
• Outperformed three state-of-the-art algorithms (HMOEA/D, MOEA/DD, NSGAIII) on 37 benchmark datasets.
• Effectively balances the goals of Pareto optimality and solution diversity.

Conclusions:

• The collaborative neurodynamic approach is a promising method for multiobjective optimization.
• The proposed technique offers significant advantages in achieving both solution quality and diversity.
• Provides a robust framework for characterizing the efficient frontier in complex optimization problems.