Then, considering a problem of pattern discrimination, the
vector z is made up of m principal discrimination components.
Imagining that these components develop in a hierarchy form,
the first component is formed by orthogonal projection of the
vector x in the direction w1 ðlw1 l ¼ 1Þ; which represents the
principal direction of the discrimination of the data in
question (Fig. 1).
Then the principal component p1 is the orthogonal
projection of x in w1 and can be represented as