Page 109 - DCAP506_ARTIFICIAL_INTELLIGENCE

Unit 8: Statistical Reasoning

We will gaze at wide categories: Notes
 Certainty factors,
 Bayesian networks.

Task Discuss the problems occurring in bayes theorem.

Self Assessment

Fill in the blanks:
1. ....................... methods give a method for showing principles that are not certain (or
uncertain) but for which there may be some assisting (or contradictory) confirmation.
2. ....................... are (real) numbers in the range 0 to 1.
3. ....................... probability, P(A|B), signifies the probability of event A specified that we
know event B has appeared.
4. The set of all ....................... must be mutually exclusive and comprehensive.
5. P(.......................) specifies the probability of A specified only B’s evidence.
6. Probability = (number of desired outcomes) / (.......................)

8.2 Certainty Factors and Rule Based Systems

This strategy has been recommended by Shortliffe and Buchanan and utilized in their famous
medical diagnosis MYCIN system.
MYCIN is fundamentally and expert system. Here we only focus on the probabilistic reasoning
aspects of MYCIN.
 MYCIN signifies knowledge as a set of rules.

 Related with each rule is a certainty factor
 A certainty factor depends on measures of belief B and disbelief D of an hypothesis H
i
given evidence E as below:
1
  max[P(H i |E),P(H )]

B(H |E)   i if P(H ) 1 otherwise
i
i

  (1 P(H ))P(H |E)
i
i
1
  P(H i ) min[P(H |E),P(H )]

D(H |E)   i i if P(H ) 0 otherwise
i
i
  P(H ))P(H |E)
i
i
where P(H ) is the standard probability.
i
 The certainty factor C of some hypothesis H given evidence E is defined as:
i
C(H |E) = B(H |E) – D(H |E)
i i i
LOVELY PROFESSIONAL UNIVERSITY 103