| Description |
If a learning algorithm uses an attribute-value representation, each
instance of the underlying universe of discourse is described by
the values of a given set of attributes, as common in relational
databases.
Each attribute can be regarded as a function, mapping the
universe of discourse to a specific domain.
If we say the example language of a learning algorithm is an
attribute-value representation, without specifying the type of
attributes, we usually mean, that the domain of each attribute is
finite or nominal and each value is treated by the
learning algorithm without using a specific interpretation.
In contrast, if we talk about numerical attributes, we often think of
attributes having infinite domains with a natural interpretation a
learning algorithm may use.
|
