latest version v1.9 - last update 10 Apr 2010 |
Unsupervised classifiers are methods that try to find 'natural' or 'sensible' structure in data.
To this end each data point is usually member of a newly found group. A distinct label is assigned to each group. Some algorithms allow membership of each point to more then one group. One example is the Fuzzy-C-Means Clustering.
The most popular application of unsupervised classifiers is data analysis and possibly visualization. However, some methods are also used for data compression or quantization.
This page first deals with clustering techniques. These are usually statistical methods for finding groups (clusters). The second section deals with an artificial neural network, the Self-Organizing Feature Maps, which perform unsupervised classification.
Related classes:
Still to be done.