A novel algorithm with differential evolution and coral reef optimization for extreme learning machine training.

Area of Science:

• Machine Learning
• Computational Intelligence
• Artificial Neural Networks

Background:

• Extreme Learning Machine (ELM) offers fast training for single-layer feed-forward networks.
• Increased hidden neurons in ELM can reduce prediction speed.
• Existing methods like Differential Evolution (DE) for ELM face local optima issues.

Purpose of the Study:

• To propose a novel hybrid algorithm, Differential Evolution Coral Reef Optimization (DECRO), for ELM training.
• To balance exploration and exploitation for improved ELM performance.
• To reduce prediction time and enhance accuracy in ELM.

Main Methods:

• Hybridization of Differential Evolution (DE) and Coral Reef Optimization (CRO) into DECRO.
• Application of DE, CRO, and DECRO algorithms to train ELM.
• Comparative experimental analysis of DECRO-ELM against DE-ELM and CRO-ELM.

Main Results:

• DECRO-ELM demonstrates reduced prediction time compared to the original ELM.
• The proposed DECRO algorithm achieves superior performance in ELM training over DE and CRO.
• The hybrid approach effectively balances exploration and exploitation for better optimization.

Conclusions:

• DECRO-ELM offers a significant improvement in both speed and performance for training Extreme Learning Machines.
• The DECRO algorithm provides a robust solution to overcome limitations of DE and CRO in ELM optimization.
• This hybrid metaheuristic approach holds promise for advancing machine learning model training.