MATLAB Source Code for Incremental Extreme Learning Machine

This MATLAB source code provides a complete implementation of the Incremental Extreme Learning Machine (I-ELM) algorithm, featuring a well-organized codebase suitable for academic research and practical applications. The implementation includes core functionalities such as random weight initialization, incremental hidden node addition, and output weight calculation using Moore-Penrose pseudoinverse. The code demonstrates efficient handling of sequential learning tasks where new training data arrives incrementally, making it particularly valuable for real-time machine learning applications. Key algorithmic components implemented include: - Dynamic network architecture expansion through sequential hidden node addition - Efficient recursive weight update mechanism avoiding full retraining - Activation function handling (sigmoid, RBF, or other kernel functions) - Performance evaluation metrics for model validation The code structure follows MATLAB best practices with clear function separation, making it easy to modify and extend for specific research needs. Undergraduate students can utilize this implementation to understand fundamental concepts of online learning, neural network optimization, and adaptive system design. The commented code provides insights into parameter tuning, convergence analysis, and computational efficiency considerations essential for machine learning projects. This resource serves as both a practical tool for immediate application and an educational platform for studying advanced machine learning techniques, particularly in domains requiring incremental model updates and adaptive learning capabilities.