Page 118 - Introduction to Statistical Pattern Recognition
P. 118
100 Introduction to Statistical Pattern Recognition
1.0 7
difference. It is important to know which term is dominant, because it deter-
mines what type of a classifier must be designed for given distributions.
Example 12: Figure 3-18 shows the Chernoff distance p(s) for DATA
I-A. Although the two covariance matrices are significantly different in this
case, the optimum s is so = 0.58, which is close to 0.5. Also, the resulting
bound of E, = 0.5e-pL(S") = 0.046 (assuming P1 = P2 = 0.5) is very close to the
