Modified Marine Predators Algorithm hybridized with teaching-learning mechanism for solving optimization problems.

Area of Science:

• Optimization Algorithms
• Nature-Inspired Computing
• Computational Intelligence

Background:

• The Marine Predators Algorithm (MPA) is a nature-inspired meta-heuristic algorithm based on predator-prey dynamics.
• Existing MPA versions suffer from easily falling into local optima and premature convergence.
• Balancing exploration and exploitation is crucial for meta-heuristic algorithm performance.

Purpose of the Study:

• To introduce a modified Marine Predators Algorithm hybridized with a teaching-learning mechanism, named MTLMPA.
• To address the shortcomings of the original MPA, specifically local optima and premature convergence.
• To enhance both the exploitation and exploration capabilities of the MPA.

Main Methods:

• Hybridization of the Marine Predators Algorithm (MPA) with a teaching-learning mechanism.
• Introduction of a teaching mechanism in the first phase to improve global search.
• Incorporation of a novel learning mechanism in the third phase to increase predator-prey encounter rates and avoid premature convergence.

Main Results:

• The proposed MTLMPA demonstrated improved global searching ability.
• The MTLMPA effectively avoided premature convergence and enhanced exploration.
• Performance was validated using 23 benchmark numerical functions and 29 CEC-2017 test functions.

Conclusions:

• The MTLMPA offers a significant improvement over the standard MPA.
• The hybridized approach effectively balances exploitation and exploration in optimization.
• MTLMPA proves to be a competitive and effective optimization algorithm compared to state-of-the-art methods.