2.1.3 Classification and Clustering

In the context of neural networks, classification involves deriving a function that will separate data into categories, or classes, characterized by a distinct set of features. This function is mechanized by a so-called network classifier, which is trained using data from the different classes as inputs, and vectors indicating the true class as outputs.

A network classifier typically maps a given input vector to one of a number of classes represented by an equal number of outputs, by producing 1 at the output class and 0 elsewhere. However, the outputs are not always binary (0 or 1); sometimes they may range over {0,1}, indicating the degrees of participation of a given input over the output classes. The Neural Networks package contains some functions especially suited for this kind of constrained approximation.

The following types of neural networks are available for solving classification problems:

Perceptron

Vector Quantization (VQ) Networks

Feedforward Neural Networks

Radial Basis Function Networks

Hopfield Networks

A basic classification example can be found in Section 3.4.1.

When the desired outputs are not specified, a neural network can only operate on input data. As such, the neural network cannot be trained to produce a desired output in a supervised way, but must instead look for hidden structures in the input data without supervision, employing so-called self-organizing. Structures in data manifest themselves as constellations of clusters that imply levels of correlation among the raw data and a consequent reduction in dimensionality and increased information in coding efficiency. Specifically, a particular input data vector that falls within a given cluster could be represented by its unique centroid within some squared error. As such, unsupervised networks may be viewed as classifiers, where the classes are the discovered clusters.

An unsupervised network can also employ a neighbor feature so that "proximity" among clusters may be preserved in the clustering process. Such networks, known as self-organizing maps or Kohonen networks, may be interpreted loosely as being nonlinear projections of the original data onto a one- or two-dimensional space.

Unsupervised networks and self-organizing maps are described in some detail in section Section 2.8, Unsupervised and Vector Quantization (VQ) Networks.