A name for kernel density estimation once common in the pattern
A simple classifier into two classes which thresholds a linear
combination of the features. Much publicized by F. Rosenblatt
A classifier constructed by assuming that estimated parameter
values are in fact the true ones.
The probability of an event conditional on the observations.
A classifier constructed by averaging over the uncertainty in
the estimated parameter values.
are linear combinations of features with high variance.
Probabilities specified before seeing the data, and so based
on prior experience or belief. Commonly these are the prior
probabilities of the classes.
Suppose we divide the parameters The profile likelihood for
is the likelihood for maximized over .
methods are based on extracting features (linear combinations
of the original features.) Exploratory projection pursuit (Section
9.1) looks for 'interesting' (non-normal) features, and projection
pursuit regression uses the extracted features in an additive
is the term used for removing parts of trees and networks with
the aim of increasing generalization.