Self-Organizing Map (SOM) Implementation in Python

This repository contains an implementation of a Self-Organizing Map (SOM), a type of artificial neural network that is trained using unsupervised learning to produce a low-dimensional (typically two-dimensional) representation of the input space. This implementation is particularly useful for visualizing high-dimensional data in a lower-dimensional space.

Features

• Initialization Modes: Diagonal, One Point, Random.
• Adjustable Parameters: Learning rate, neighborhood influence, and number of neurons.
• Dynamic Neighborhood Updating: Automatic updating during training.
• Visualization Support: Visualize the training process in real-time.
• Data Mapping: Mapping of data points to neurons for cluster identification.

Dependencies

• numpy
• matplotlib
• sklearn (for example data generation)

Installation

No specific installation steps are required apart from ensuring the required dependencies are installed. You can install the dependencies using pip:

pip install numpy matplotlib scikit-learn

Usage

Step 1: Import the SOM class and Auxiliary Functions

from som_module import SOM, analyze_array

Step 2: Prepare Your Data

from sklearn.datasets import make_blobs
data, _ = make_blobs(n_samples=1000, centers=50, n_features=7, random_state=42)

Step 3: Initialize the SOM

num_neurons = int(np.ceil(np.sqrt(data.shape[0])))
epochs = 100
learning_rate = 0.3
update_neighbors_epoch = 5
influence = 0.1
calculate_k_epoch = 2
k_neighbors = 5

som = SOM(data, num_neurons, epochs, learning_rate, influence, update_neighbors_epoch, calculate_k_epoch, k_neighbors, randomize_data=False, init_mode='diagonal')

Step 4: Train the SOM

mapped_data = som.train(step_by_step=True)

Step 5: Analyze the Results

print(analyze_array(mapped_data))

Step 6: Visualizing the SOM (Optional)

Set step_by_step to True to enable real-time training visualization.