Page 217 - Introduction to Statistical Pattern Recognition

5 Parameter Estimation 199

(# of ol-errors) (# of 02-errors)
&=PI 4- P2 (5.44)
N, N2

A
That is, E of (5.40) is the error obtained by counting the number of
misclassified samples with a given classifier.

Moments of i: The expected value of aYi with respect to Xy' (w.r.t. the
test sample) is

(5.45)

The second line of (5.45) can be obtained from (5.35) and (5.36) respectively.
The second order moments are also computed as

1
-_ (5.46)
-
4'

Equation (5.47) is obtained because a?' and are independent due to the
independence between Xy) and Xlkj.