Page 110 - DCAP506_ARTIFICIAL_INTELLIGENCE

Artificial Intelligence

Notes 8.2.1 Reasoning with Certainty Factors

 Rules articulated as if evidence list E , E then there is indicative evidence with probability,
1 2
p for symptom H .
i
 MYCIN accesses rules to reason backward to clinical data evidence from its aim of
forecasting a disease-causing organism.
 Certainty factors primarily supplied by experts changed as per the previous formulae.

 How do we perform reasoning when numerous rules are chained together?
Measures of belief and disbelief given numerous observations are computed as follows:


B(H |E , E )  0 if D(H |E ) 1 otherwise
i 1 2  B(H |E )  B(H i |E )(1 B(H |E )) i 1 i 2
i
2
1
D(H |E)  0 if B(H |E ) 1 otherwise

i  D(H |E )  D(H i |E )  D(H i |E )) 1 2
1
i
1
2
 How about our belief regarding several hypotheses taken jointly? Measures of belief
given numerous hypotheses and to be shared logically are computed as follows:
B(H  H |E)  min(B(H |E) B(H |E)
2
1
1
1
1
B(H  H |E)  max(B(H |E) B(H |E)
1 2 1 1 1
Disbelief is computed similarly.
8.2.2 Overcoming the Bayes Rule Shortcomings
Certainty Factors do hold on to the rules of Bayesian statistics, but it can symbolize tractable
knowledge systems:
 Individual rules give belief in an hypotheses — fundamentally a conditional probability.
 The formulae for amalgamation of evidence/hypotheses fundamentally assume that all
rules are sovereign ruling out the requirement for joint probabilities.
 The load of assuring independence is positioned on the rule writer.

Self Assessment

Fill in the blanks:
7. MYCIN signifies knowledge as a set of ....................... .
8. MYCIN accesses rules to reason backward to clinical data evidence from its aim of
forecasting a ....................... causing organism.
9. The load of assuring ....................... is positioned on the rule writer.

8.3 Bayesian Networks

These are also known as Belief Networks or Probabilistic Inference Networks primarily generated by
Pearl (1988).
The fundamental idea is:

 Knowledge in the world is modular — many events are conditionally sovereign of most
other events.

104 LOVELY PROFESSIONAL UNIVERSITY