The self-organizing map (SOM), as a kind of unsu-
pervised neural network, is performed in a selforganized
manner in that no external teacher or
critic is required to guide synaptic changes in the
network.
By contrast, for the other two basic
learning paradigms in neural networks, supervised
learning is performed under the supervision of an
4,13
external teacher
and reinforcement learning involves
the use of a critic that evolves through a trialand-error
process
8
; these other two also demand
the input-output pairs as the training data. The
appealing features of learning without needing the
input-output pairs makes the SOM very attractive
when dealing with varying and uncertain data. In
its many applications, the SOM has been used for
both static data management and dynamic data
analysis, such as data mining, knowledge discovery,