Discriminant Functions

鉴别函数(Discriminant Functions)

	设定鉴别函数概念的目的是将归类的判断条件统一化
  1. set-of discriminant functions g i ( x ) , i = 1 , 2 , … , c g_i(x),i=1,2,\ldots,c gi(x),i=1,2,,c.
  2. if g i ( x ) = max ⁡ g j ( x ) g_i(x) = \max g_j(x) gi(x)=maxgj(x),then x → w i x \to w_i xwi (or if g i ( x ) − g j ( x ) > 0 g_i(x)-g_j(x)>0 gi(x)gj(x)>0, then x → w i x\to w_i xwi)

Example:

比如对于最小错误率贝叶斯公式,我们可以让鉴别函数为如下:

  1. g i ( x ) = P ( w i ∣ x ) g_i(x)=P(w_i|x) gi(x)=P(wix)
  2. g i ( x ) = P ( w i ) P ( x ∣ w i ) g_i(x)=P(w_i)P(x|w_i) gi(x)=P(wi)P(xwi)
  3. g i ( x ) = l n P ( w i ) + l n P ( x ∣ w i ) g_i(x)=lnP(w_i)+lnP(x|w_i) gi(x)=lnP(wi)+lnP(xwi)

特别的,对于只有两个类别的问题,可以采用如下方式:

g ( x ) = g 1 ( x ) − g 2 ( x ) g(x)=g_1(x)-g_2(x) g(x)=g1(x)g2(x)
if g ( x ) > 0 g(x)>0 g(x)>0;
x → w 1 x\to w_1 xw1
else
x → w 2 x \to w_2 xw2

其中 g ( x ) g(x) g(x) 可以是如下:

  1. g ( x ) = P ( w 1 ∣ x ) − P ( w 2 ∣ x ) g(x) = P(w_1|x)-P(w_2|x) g(x)=P(w1x)P(w2x)
  2. g ( x ) = l n P ( w 1 ∣ x ) − l n P ( w 2 ∣ x ) g(x)=lnP(w_1|x)-lnP(w_2|x) g(x)=lnP(w1x)lnP(w2x)
  3. g ( x ) = P ( x ∣ w 1 ) P ( w 1 ) − P ( x ∣ w 2 ) P ( w 2 ) g(x)=P(x|w_1)P(w_1)-P(x|w_2)P(w_2) g(x)=P(xw1)P(w1)P(xw2)P(w2)
  4. g ( x ) = l n P ( x ∣ w 1 ) P ( x ∣ w 2 ) + l n P ( w 1 ) P ( w 2 ) g(x)=ln\frac {P(x|w_1)}{P(x|w_2)}+ln\frac{P(w_1)}{P(w_2)} g(x)=lnP(xw2)P(xw1)+lnP(w2)P(w1)

你可能感兴趣的:(模式识别)