Combining Symbolic and Neural Learning to Revise Probabilistic Rule Bases
J. Jeffrey Mahoney · 1992
This paper describes RAPTURE -- a system for revising probabilistic rule bases that combines symbolic and neural-network learning methods. RAPTURE uses a modified version of back- propagation to refine the certainty factors of a MYCIN-style rule base and it uses ID''s information gain heuristic to add new rules. Current results on two real-world domains are presented, demonstrating that this combined approach performs as well or better than previous methods. Future work for this project is discussed, which includes further testing on other domains, as well as experimentation with several network training techniques. Possible extensions to this project include using different probabilistic formalisms for the rule base such as Bayes Nets, Dempster- Schafer theory, and Fuzzy-Logic.