| INCR/CSL is a system for the incremental learning of rules from examples adopting a closed loop architecture. The underlying representation language is Michalski's VL21. It is able to refine a starting theory generated by the batch learning system INDUBI/CSL, by assimilating new knowledge expressed as ground clauses. The refinement is performed by upward and downward refinement operators, as well as an abstraction operator performing saturation. | | |