An Efficient Hybrid Rule Based Inference Engine with Explanation Capability

Aninference enginefor a hybrid representation scheme basedon neurules is presented. Neurulesare a kind of hybridrules that combinea symbolic(productionrules) and a connectionist representation (adaline unit). The inferenceengineusesa connectionisttechnique,whichis basedonthe 'firing potential', a measurementof the firing tendencyof a neurule,andsymbolicpatternmatching.It is proved to be moreefficient and natural than pure connectionistinferenceengines.Explanationof 'how'type canbeprovidedin the formof if-then symbolicrules.

• Type: Conference
• Year: 2001
• Where: FLAIRS
• Authors: Ioannis Hatzilygeroudis, Jim Prentzas
• Keywords: Artificial Intelligence, Connectionist Representation, FLAIRS 2001, Hybrid Representation Scheme, Inferenceengineusesa Connectionisttechnique