Understanding 2D Self-Organising Maps: A Journey Through…

What is a 2D Self-Organising Map

A 2D self-organising map (SOM) is a type of artificial neural network that maps input data onto a two-dimensional grid. Each node in the grid, known as a neuron, represents a prototype vector that corresponds to a region in the input space.

The SOM algorithm organizes these neurons based on their similarity to the input data points through competitive learning and topological preservation, allowing for a visual representation of high-dimensional data.

How It Works

During training, each neuron in the grid is associated with a weight vector. The algorithm starts by randomly initializing these vectors. For each input sample, the neuron whose weight vector is closest to the input (the Best Matching Unit or BMU) and its neighbors are updated.

The updating process involves adjusting the weights of the BMU and its neighbors towards the input data point. Over time, this process leads to a grid where similar inputs cluster together in close proximity.

live demo · related simulation● LIVE

Why It Matters

Self-organising maps are valuable for visualizing complex datasets because they provide an intuitive way to understand the structure and relationships within the data. They can reveal patterns, clusters, and anomalies that might not be apparent in raw data.

In applications such as image processing, bioinformatics, and market research, SOMs help in reducing dimensionality while preserving meaningful information.

Real-World Applications

SOMs are used in a variety of fields. In neuroscience, they can model the organization of sensory maps in the brain. In data mining and pattern recognition, SOMs help in clustering large datasets into meaningful groups.

In telecommunications, SOMs assist in optimizing network configurations by identifying key areas for resource allocation.

Frequently asked questions

How does a self-organising map differ from other clustering methods?

Unlike traditional clustering algorithms that group data points based on distance metrics, SOMs preserve the topological structure of the input space. This means that similar data points are not only grouped together but also remain close to each other in the grid.

Can self-organising maps be used for more than just clustering?

Yes, while SOMs are primarily known for clustering and visualization, they can also be used for function approximation, vector quantization, and even as a pre-processing step in other machine learning algorithms.

What is the role of the neighbourhood update in self-organising maps?

The neighbourhood update mechanism ensures that not only the BMU but also its neighbors are updated. This helps to maintain the topological structure of the map, ensuring that similar data points remain close to each other.

How can I choose the appropriate grid size for my self-organising map?

The choice of grid size depends on the complexity and dimensionality of your dataset. A larger grid provides finer resolution but requires more computational resources. Typically, a good starting point is to use a square grid with 10 to 20 neurons per side.

Try it live

Everything above runs in your browser — open 2D Self-Organising Map and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.

▶ Open 2D Self-Organising Map simulation

Understanding 2D Self-Organising Maps: A Journey Through Machine Learning