Data Structure
Networking
RDBMS
Operating System
Java
MS Excel
iOS
HTML
CSS
Android
Python
C Programming
C++
C#
MongoDB
MySQL
Javascript
PHP
Physics
Chemistry
Biology
Mathematics
English
Economics
Psychology
Social Studies
Fashion Studies
Legal Studies
- Selected Reading
- UPSC IAS Exams Notes
- Developer's Best Practices
- Questions and Answers
- Effective Resume Writing
- HR Interview Questions
- Computer Glossary
- Who is Who
What is SOM?
SOM represents Self-Organizing Feature Map. It is a clustering and data visualization approaches depends on a neural network viewpoint. The objective of SOM is to discover a set of centroids (reference vectors in SOM terminology) and to create each object in the data set to the centroid that supports the best closeness of that object. In neural network methods, there is one neuron related to each centroid.
As with incremental K-means, data objects are phased one at a time and the nearest centroid is refreshed. Unlike K-means, SOM imposes a topographic sequencing on the centroids and nearby centroids are also upgraded. Moreover, SOM does not keep mark of the recent cluster membership of an object, and, unlike K-means, if an object switches clusters, there is no specific refresh of the old cluster centroid.
The old cluster can be in the neighborhood of the new cluster and therefore cab be updated for that reason. The processing of points continues until some pre-decided limit is reached or the centroids are not transforming very much. The last output of the SOM approaches is a set of centroids that implicitly represent clusters. Each cluster includes the points closest to a specific centroid.
Each centroid is created a pair of coordinates (i, j). Sometimes, such a network is drawn with connection among adjacent nodes, but that can be misleading because the power of one centroid on another is a neighborhood that is represented in method of coordinates, not links. There are several types of SOM neural networks, but it can confine this discussion to two-dimensional SOMs with a rectangular or hexagonal organization of the centroids.
Centroids used in SOM have a pre-decided topographic sequencing relationship. During the training procedure, SOM needs each data point to refresh the nearest centroid and centroids that are nearby in the topographic sequencing. In this method, SOM produces an ordered set of centroids for any given data set.
In another terms, the centroids that are near to each other in the SOM grid are more closely associated to each other than to the centroids that are further away. Because of this constraint, the centroids of a two-dimensional SOM can be considered as lying on a two-dimensional surface that attempts to fit the n-dimensional data further possible.
195 Views
