Feature extraction is the transformation of a raw observation 
into a feature vector, an
ordered list of numerical or categorical measurements that describes the observation
and is more useful for analysis or classification.
It can be represented as a function
 |
where 
is the observation space and 
is the feature space. The transformation is chosen to retain
information relevant to the task while suppressing irrelevant variation or reducing
dimension. For example, a linear transformation
can project data onto the leading directions found by principal
component analysis.
