Kohonen's SOFM (Self-Organizing Feature Map) Algorithm Implementation

This repository contains the source code implementation of Kohonen's SOFM (Self-Organizing Feature Map) algorithm. The algorithm employs unsupervised learning to map high-dimensional data into lower-dimensional spaces while preserving topological properties of the original data. This implementation features a competitive learning mechanism where neurons in the output layer compete to respond to input patterns, with the winning neuron and its neighbors updating their weights through a neighborhood function. Key implementation aspects include: - Weight initialization using random values or principal components - Neighborhood function implementation (typically Gaussian or bubble) - Learning rate decay strategies over training epochs - Distance calculation methods (Euclidean, Manhattan) for best matching unit selection - Topology preservation through grid structures (rectangular or hexagonal) The algorithm is particularly valuable in data mining and machine learning for applications such as classification, clustering, and data visualization. The name honors its inventor Teuvo Kohonen, a prominent Finnish mathematician and computer scientist. This implementation includes configurable parameters for grid size, learning rates, and iteration counts to facilitate experimentation with different datasets. The code structure features modular components for: - Data preprocessing and normalization - Network initialization and training loops - Visualization utilities for mapping results - Performance evaluation metrics This source code provides a solid foundation for understanding and applying self-organizing maps in practical machine learning scenarios.