Computationally Tractable, Conceptually Plausible Classes of Link Matrices for the Inquery Inference Network
MASSACHUSETTS UNIV AMHERST DEPT OF COMPUTER SCIENCE
Pagination or Media Count:
This paper presents the PIC matrices, a computationally efficient subclass of link matrices that may be considered for the interpretation of query operators in the INQUERY inference network. The PIC class of matrices is formally defined a defined 0nexp 2 algorithm, PIC-EVAL, is presented for the evaluation of members of this class and a proof of the correctness of the algorithm is given. A further specalization of this class that can be evaluated even more efficiently is discussed. Finally, a generalization of the PIC class, for which the input probabilities may be viewed as weighted, is defined, and a simple modification of the PIC-EVAL algorithm that allows for the evaluation of matrices of this extended class is given.
- Information Science
- Computer Programming and Software
- Radio Communications