Success History Adaptive Competitive Swarm Optimizer with Linear Population Reduction: Performance benchmarking and...

Area of Science:

• Artificial Intelligence
• Computational Optimization
• Medical Diagnostics

Background:

• Artificial intelligence (AI) has advanced eye disease detection, but model accuracy is limited by optimizer deficiencies.
• Existing optimizers struggle to balance exploration and exploitation effectively during complex optimization tasks.

Purpose of the Study:

• To introduce an efficient optimizer, Success History Adaptive Competitive Swarm Optimizer with Linear Population Reduction (L-SHACSO), to enhance AI model performance.
• To evaluate the superiority of L-SHACSO against state-of-the-art optimizers on benchmark and engineering problems.
• To apply L-SHACSO for improving the accuracy of eye disease detection models.

Main Methods:

• Developed L-SHACSO by integrating success history adaptation and linear population reduction strategies into the Competitive Swarm Optimizer (CSO).
• Conducted extensive numerical experiments on CEC2017, CEC2020, CEC2022, and eight engineering problems, comparing L-SHACSO with jSO, L-SHADE-cnEpSin, RIME, and Parrot Optimizer (PO).
• Integrated L-SHACSO with DenseNet and Extreme Learning Machine (ELM) to create the DenseNet-L-SHACSO-ELM model for eye disease classification.

Main Results:

• L-SHACSO demonstrated superior performance across various optimization tasks compared to competing state-of-the-art algorithms.
• The DenseNet-L-SHACSO-ELM model achieved high accuracy in eye disease detection on public datasets.
• The proposed model confirmed the feasibility and effectiveness of L-SHACSO in a practical medical application.

Conclusions:

• L-SHACSO is an efficient optimizer that significantly enhances AI model performance, particularly in complex tasks like medical image analysis.
• The integration of L-SHACSO into deep learning models offers a promising approach for accurate and reliable eye disease detection.
• The developed L-SHACSO optimizer has substantial potential for real-world applications in healthcare and beyond.