Page 409 - Introduction to Information Optics

7. Pattern Recognition with Optics

0 2 4 6 8 10 12 14 16 18
Quantization Level
(a)

2
4 6 8 10 12 14
Quantization Level

Fig. 7.33. (a) Discriminability and (b) reliability as a function of quantization level.

bility performs better for higher quantization levels, and it decreases rather
rapidly as the standard deviation of the noise increases. Also note that, as the
quantization level increases, the improvement of the FR tends to level off, as
shown in Fig. 7.33a.

7.6. PATTERN CLASSIFICATION

Artificial neural pattern classifiers can be roughly classified into four groups:
global discriminant classifiers, local discriminant classifiers, nearest neighbor
classifiers, and rule-forming classifiers. In this section, we focus on nearest
neighbor classifiers (NNCs) that perform classifications based on the distance