Functional extreme learning machine for regression and classification.

Area of Science:

• Artificial Intelligence
• Machine Learning
• Neural Networks

Background:

• Traditional Extreme Learning Machines (ELM) offer rapid training but suffer from limited fitting accuracy.
• Gradient-based algorithms for neural network training are often slow and computationally intensive.

Purpose of the Study:

• To introduce Functional Extreme Learning Machine (FELM), a novel regression and classification model.
• To enhance the accuracy and generalization capabilities of ELM while maintaining its speed.

Main Methods:

• FELM utilizes functional neurons as its core computational units.
• The model employs functional equation-solving theory for its development.
• It determines hidden layer coefficients by solving the generalized inverse of the output matrix, avoiding iterative optimization.

Main Results:

• FELM demonstrates comparable learning speed to ELM.
• Experimental results indicate superior generalization performance and stability for FELM compared to ELM.
• FELM outperformed ELM, OP-ELM, SVM, and LSSVM on various synthetic, XOR, and benchmark datasets.

Conclusions:

• FELM represents a significant advancement over traditional ELM for regression and classification tasks.
• The functional neuron approach combined with equation-solving theory leads to improved model accuracy and robustness.
• FELM offers a viable alternative for applications requiring fast yet accurate neural network models.