Experimental Results on the Efficiency of Rule-Based Systems
In rule-based systems several strategies can be developed which permit to speed up the inference process, always leaving the quality of the results unchanged. For this purpose, the set of all rules is partitioned into subsets according to one or another heuristic principle; in the beginning, the inference process is run on a small subset only, whereas later on further subsets of rules are activated step by step, until finally the total set of rules is included. Experimental results on the speed-up, which depends on the way of partitioning the set of rules, will be presented, together with some further data on the parameterdependency of the time required for inference processes.
Unable to display preview. Download preview PDF.