Efficient Forward Chaining for Declarative Rules in a Multi-Agent Modelling Language
Section 4 describes how SDML fires rules using forward and backward chaining. Rulebases are typically regarded as unstructured collections of rules, but they contain implicit structure, as defined by the dependencies between rules. In addition to helping users construct rulebases and detect errors, they are extremely useful for efficient forward chaining.
Section 5 describes how SDML can make provisional deductions that may be invalidated by new information. This is commonly referred to as non-monotonic reasoning [9]. If a deduction is made assuming that a predicate is not true, and this predicate is later deduced to be true, then the deduction must be retracted. It would be very restrictive if the deduction could only have been made if the predicate had already been proven to be false.
Section 6 provides some measurements of the speed of SDML, and compares it with an imperative rule-based system. Section 7 compares the features of SDML with other systems. Finally, Section 8 presents some conclusions.
Generated with CERN WebMaker