Page 229 - Introduction to Statistical Pattern Recognition
P. 229
5 Parameter Estimation 21 1
(5.89)
Substituting (5.82)-(5.89) into (5.70), v of (5.71) becomes
(5.90)
where
Since ,f,(X,w) is a linear combination of xf. (k I2), v,- can be theoretically
computed for Data I-I, resulting in
(5.92)
Equation (5.92) was experimentally verified in the same manner as
(5.81). The results are shown in Table 5-7.
Experiment 5: Error of the linear classifier
Data: 1-1 (Normal, MTM = 2.562, E = 10%)
Dimensionality: n = 4, 8, 16, 32, 64
Classifier: Linear classifier of (5.55)
Design samples: YI = y.2 = kn, k = 3, 5, 10, 20, 40
Tcst: Theoretical using the normal error table
No. of trials: z = 10
Results: Table 5-7 [4]
