An Efficient Leave-One-Out Cross-Validation-Based Extreme Learning Machine (ELOO-ELM) With Minimal User Intervention

Area of Science:

• Machine Learning
• Computational Intelligence
• Neural Networks

Background:

• Extreme Learning Machine (ELM) architecture is crucial for performance.
• Selecting optimal hidden neurons is a key challenge in ELM.
• Leave-One-Out Cross-Validation (LOO-CV) provides reliable model selection but is computationally expensive.

Purpose of the Study:

• To develop an efficient LOO-CV formula and algorithm for ELM.
• To enable fast learning speeds comparable to standard ELM.
• To maintain the reliability of LOO-CV while improving efficiency.

Main Methods:

• Proposed an efficient LOO-CV formula.
• Developed an efficient LOO-CV-based ELM (ELOO-ELM) algorithm.
• Validated the algorithm on benchmark datasets.

Main Results:

• The ELOO-ELM algorithm achieves fast learning speeds.
• The algorithm retains the reliability of LOO-CV.
• Minimal user intervention is required, facilitating adoption by non-experts and automation.

Conclusions:

• The ELOO-ELM algorithm offers an efficient and reliable method for ELM model selection.
• It achieves good generalization performance with limited user input.
• The approach enhances the practical applicability of ELM in various domains.