The triggering and control of inference processes (ficsr)
James Steven Botic · 1983
This research examines the problems associated with computer implementation of inference processes. We investigate two questions: (1) How can inference triggering be controlled? and (2) How can the depth of inference process(es) be controlled? To date, Pattern-Directed Inference Systems (PDIS) approach to conflict set resolution, the inclusion and ordering of pattern-directed modules (PDMs) in the conflict set, can be characterized as ad hoc or heuristic. This thesis develops a new approach to conflict set resolution combining concepts from brain theory (namely the neuron) and fuzzy logic (the membership function). The combination of these concepts results in a mechanism, Fuzzy Interest Conflict Set Resolution (FICSR), that controls the inclusion of PDMs in the conflict set via a membership function, u(,I)(PDM), and threshold function t(,I)(depth), each being determined, to some extent, by a current interest set I. The threshold function not only controls inclusion of PDMs into the conflict set, but the depth of the inference process as well. We develop forms for u(,I)(PDM) and t(,I)(depth) and demonstrate how they resolve the conflict set resolution problem. The concepts of FICSR have been implemented in a system called FICSR. Results indicate that the combination of interests and fuzzy logic can be used to control the inference process.