不确定推理算法举例(确定因子法)
例6-1
下面通过一个例子说明该方法的推理过程。
例如有如下的推理规则:
rule 1:IF E1 THEN H(0.9)
rule 2:IF E2 THEN H(0.7)
rule 3:IF E3 THEN H(-0.8)
rule 4:IF E4 AND E5 THEN E1(0.7)
rule 5:IF E6 AND (E7 OR E8) THEN
E2(1.0)
形成的推理网络如图5-2所示。
图5-2 规则形成的推理网络
在图5-2中,E3、E4、E5、E6 、E7和E8为原始证据,其确定性因子由用户给出,假定它们的值为:
cF(E3)=0.3 cF(E4)=0.9
cF(E5)=0.6
cF(E6)=0.7 cF(E7)=-0.3
cF(E8)=0.8
系统的求值顺序是先求出cF(E1)、cF(E2)和cF(E3),然后再求cF(H)。
cF(E1)=0.7*max{0,cF(E4
AND E5)}
=0.7*max{0,min{cF(E4),cF(E5)}}
=0.7*max{0,min{0.9,0.6}}
=0.7*0.6
=0.42
cF(E2)=1*max{0,cF(E6
AND(E7 OR E8))}
=1*max{0,min{cF(E6),max{cF(E7),cF(E8)}}}
=1*max{0,min{cF(E6),max{-0.3,0.8}}}
=1*max{0,min{0.7,0.8}}
=1*0.7
=0.7
cF(E3)=0.8
cF1(H)=0.9*max{0,cF(E1)}
=0.9*max{0,0.42}
=0.9*0.42
=0.38
cF2(H)=0.7*max{0,cF(E2)}
=0.7*max{0,0.7}
=0.7*0.7
=0.49
cF3(H)=-0.8*cF(E3)
=-0.8*0.3
=-0.24
因为cF1(H)>0且cF2(H)>0
所以cF12=cF1(H)+cF2(H)-cF1(H)*cF2(H)
=0.38+0.49-0.38*0.49=0.6838
又因为cF1(H)>0,cF2(H)<0
所以cF123=cF(H)=cF12(H)+cF3(H)
=0.6838-0.24=0.4438
所以由此推出结论H的确定性因子cF(H)=0.4438。
习题
设某问题求解用到以下的推理规则:
rule1: IF E1 THEN H1(0.9)
rule2: IF E2 THEN H1(0.5)
rule3: IF E3 and E4 THEN E1(0.8)
rule4: IF E5 and E6 THEN E2(1.0)
试画出其推理网络图,用确定因子法求出H1的可信度cF(H1,E1&E2),假定已知在当前观察有
cF(E3)=0.3,cF(E4)=0.4,cF(E5)=0.5,cF(E6)=0.6。
返回
|