Optimal dimension


Table 4.11: We applied the FLD projection to various number of dimensions, using 1-NN and nearest class mean as classifiers in projected space.
  Classification Rate [%]
Feature Space Dimension 1-NN nearest $\mathbf{\mu}$
1 41.67 35.41
2 54.17 45.83
3 64.58 50.00
4 60.42 50.00
5 72.92 66.67
6 77.08 75.00
7 77.08 75.00
8 77.17 83.33
9 79.17 83.33


Figure 4.17: The optimal feature space dimension selection for car images
\includegraphics[width=80mm,height=60mm]{fldcars.eps}



Kocurek 2007-12-17