Boyer - 23 
Perceptual Inference Network (PIN): Specification 
The conditional probabilities: 
p ( l *zr • • • 1 ^ J for i = 1, . . . , ft) = 
9^x\xi\ , • • •, tin) 
C0772p((/ Ii .,...,/ Z .),({/ Ui| ,...,/ U|fn } for i = 
( 1 ) 
where p is the conditional function which expresses the belief in 
feature .Y given the state of the other features, U \,.. •, U n and 
Comp is the locational compatibility function capturing the local 
neighborhood dependencies. The function g is of the following 
form: 
g(X = l|«i,---,u n ) = C (lili ... iUn) (l (2)