Artificial Intelligence and Soft Computing: Behavioral ... - Arteimi.info

The FPN of fig. 20.5 is formed with the rule base and database given above.
The system parameters of the FPN of fig. 20.5 are presented in Table 20.2.
Initial Weights wij
Initial Fuzzy Beliefs ni
Table 20.2: Parameters of Case History 2.
Steady-state weights after 3
iterations
thj =0 for all transitions trj
The Current Reasoning Problem
w31=0.75, w62=0.95, w43=0.8
n1=0.8, n2=0.7, n3=0.1, n4=0.2,
n5=0.9, n6=0.3
w31=0.7, w62=0.10, w43=0.10
Now, to solve a typical reasoning problem, whose knowledge and databases
are presented herewith, we need to assign the derived weights from the last
two case histories. The reasoning model can be used in this example to
compute the steady-state belief of the proposition: Hates (lata,askoke), with
the given initial beliefs of all the propositions. The system parameters of the
FPN in fig. 20.6 are presented in Table 20.3.
20.5.4 Implication of the results
The analysis of stability envisages that both the encoding and the recall model
of associative memory are unconditionally stable. The time required for
convergence of the proposed model is proportional to the number of
transitions on the largest path (cascaded set of arcs) [9] in the network. The
model could be used for determination of CF of rules in a KB by maintaining
consistency among the beliefs of the propositions of known case histories.