The design of the feature extractor is very problem dependent. The ideal
feature extractor would produce the same feature vector x
for all patterns in the same class, and different feature vectors for patterns
in different classes. In practice, different inputs to the feature extractor
will always produce different feature vectors, but we hope that the within-class
variability is small relative to the between-class variability.
At this point, we assume that the designer of the feature extractor has
done the best job he or she can, and that the feature vector contains the
information needed to distinguish the patterns. Given that feature set,
our job is to design the classifier.
Back to Vectors
Up to Models