Salp Navigation and Competitive based Parrot Optimizer (SNCPO) for efficient extreme learning machine training and...

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Machine Learning

Background:

• Metaheuristic algorithms are vital for complex problems like machine learning parameter tuning.
• Existing methods often fail to balance exploration and exploitation, causing premature convergence.
• The traditional Parrot Optimizer (PO) has limited adaptability and can get stuck in local optima.

Purpose of the Study:

• To introduce a novel hybrid optimization algorithm, the Salp Navigation and Competitive based Parrot Optimizer (SNCPO).
• To enhance the performance of metaheuristic optimization by improving the balance between exploration and exploitation.
• To address the limitations of existing algorithms in escaping local optima.

Main Methods:

• Developed SNCPO by integrating Competitive Swarm Optimization (CSO) and Salp Swarm Algorithm (SSA) into the PO framework.
• Utilized CSO's pairwise competitive learning to categorize population into winners and losers.
• Applied SSA-inspired navigation for winners (global search) and PO's strategy for losers (exploration).

Main Results:

• SNCPO demonstrated superior performance on CEC2015 and CEC2020 benchmark functions.
• The algorithm achieved better results on engineering design problems and Extreme Learning Machine (ELM) training.
• Consistent outperformance of state-of-the-art algorithms in convergence speed, solution quality, and robustness.

Conclusions:

• SNCPO effectively overcomes the limitations of traditional PO and other algorithms.
• The hybrid approach shows strong adaptability across diverse optimization landscapes.
• SNCPO holds significant potential for real-world engineering and machine learning applications.