A collaborative neurodynamic approach to global and combinatorial optimization.

Area of Science:

• Computational Intelligence
• Operations Research
• Applied Mathematics

Background:

• Combinatorial optimization problems are challenging due to their complexity.
• Existing methods may struggle with nonconvexity in objective functions or constraints.
• Global optimization requires robust algorithms to avoid local minima.

Purpose of the Study:

• To propose a novel collaborative neurodynamic optimization approach.
• To reformulate combinatorial optimization as a global optimization problem.
• To ensure convergence to globally optimal solutions for complex problems.

Main Methods:

• Reformulation of combinatorial optimization as a global optimization problem.
• Development of a neurodynamic optimization model using an augmented Lagrangian function.
• Collaborative search using multiple neurodynamic models, with initial states optimized by particle swarm optimization (PSO).

Main Results:

• The proposed neurodynamic model demonstrates asymptotic stability at strict local minima, even with nonconvexity.
• The collaborative approach effectively searches for global optimal solutions.
• The method is proven to be globally convergent and validated on benchmark problems.

Conclusions:

• The collaborative neurodynamic optimization approach offers a robust solution for global and combinatorial optimization.
• The method's stability and convergence properties are theoretically established.
• Effective application to benchmark problems confirms its practical utility.