Thanks to visit codestin.com
Credit goes to www.scribd.com

Scribd
Upload
7 views3 pages

SOM Algorithm Theory and Explanation

A Self-Organizing Map (SOM) is an unsupervised learning algorithm used for clustering and visualizing high-dimensional data by mapping it onto a lower-dimensional grid while preserving topological properties. The training process involves adjusting neuron weights based on input data to form clusters of similar inputs. SOM has applications in areas such as image compression, market segmentation, and speech recognition.

Uploaded by

siddarthabhimanyu07
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
7 views3 pages

SOM Algorithm Theory and Explanation

A Self-Organizing Map (SOM) is an unsupervised learning algorithm used for clustering and visualizing high-dimensional data by mapping it onto a lower-dimensional grid while preserving topological properties. The training process involves adjusting neuron weights based on input data to form clusters of similar inputs. SOM has applications in areas such as image compression, market segmentation, and speech recognition.

Uploaded by

siddarthabhimanyu07
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Self-Organizing Map (SOM) - Theory

A Self-Organizing Map (SOM) is an unsupervised learning algorithm developed by Teuvo Kohonen.

It is used for clustering and visualization of high-dimensional data. The SOM maps input data into a

lower-dimensional (usually 2D) grid of neurons while preserving the topological properties of the

data.

Key Concepts:

1. Each neuron (node) has a weight vector of the same dimension as the input data.

2. The neurons are arranged in a 2D grid.

3. The training process adjusts the weights so that similar inputs activate nearby neurons.

Purpose:

- Data visualization

- Pattern recognition

- Dimensionality reduction
How SOM Works - Step-by-Step (Easy Explanation)

1. Initialization:

- Each neuron in the grid is assigned a random weight vector.

2. Training Phase (Iterative):

- Randomly pick an input vector from the dataset.

- Calculate the Euclidean distance between this input and all the neuron weights.

- The neuron with the smallest distance is called the Best Matching Unit (BMU).

- Determine the BMU's neighbors in the grid.

- Update the BMU and its neighbors' weights to move closer to the input vector.

3. Updating Weights:

- The learning rate and neighborhood radius shrink over time to fine-tune the map.

4. Repetition:

- The above process is repeated for many iterations to form an organized map.

Result:

- Similar input vectors are mapped to nearby neurons, forming clusters.


SOM Components and Terminologies

Input Vector (x):

- A data sample fed into the network for training.

Weight Vector (w_ij):

- A vector associated with each neuron, adjusted during training.

Best Matching Unit (BMU):

- The neuron whose weight is most similar to the input vector.

Neighborhood Function (h_ij,b):

- A function (often Gaussian) that defines how strongly the neighbors of the BMU are affected.

Learning Rate ():

- Controls how much the weights are adjusted in each step; it decreases over time.

Topological Distance:

- The distance between neurons on the grid.

Applications:

- Image compression

- Market segmentation

- Speech recognition

- Text mining

SOM helps to organize complex data into an understandable map without needing labels.

You might also like