Page 228 - Introduction to Statistical Pattern Recognition
P. 228

210                        Introduction  to Statistical Pattern Recognition










                                                               Classifier designed by
                                                                  (fixed) samples











                          1                                        *n
                                    Fig. 5-1  An example of Hughes phenomena.



                                                                                  (5.84)




                                                                                  (5.85)


                      where

                                                                                  (5.86)

                      and El/ = (cf;)  + cf))/2  and XI = (hf” + h;”)/2.
                           In  particular,  when  h;” = I,  hj”  =IJ, rn:”  = 0,  and  rnj2) = rn, (5.84)-
                      (5.86) become


                                                                                  (5.87)



                                 a2h
                                --  -        4                                    (5.88)
                                 acy    (I+h,)(l+h/)

                      where
   223   224   225   226   227   228   229   230   231   232   233