Utility of in ecursive ain Theories

We investigate the utility of explanation-based learning in recursive domain theories and examine the cost of using macro-rules in these theories. The compilation options in a recursive domain theory range from constructing partial unwindings of the recursive rules to converting recursive rules into iterative ones. We compare these options against using appropriately ordered rules in the original domain theory and demonstrate that unless we make very strong assumptions about the nature of the distribution of future problems, it is not profitable to form recursive macro-rules via explanation-based learning in these domains.