TREE: the Heuristic Driven Join Strategy of a RETE-Like Matcher

TREE is an optimized RETE-like pattern-matching algorithm. It has been designed for a production system whose restricted data formalism leads to a highly combinatorial join step like in SOAR. TREE aims at reducing the join search spaces without using hashing techniques. Its join strategy uses constraint propagation to define the solution space of a join, then a constraint relaxation to determine the index to be used in the join computation. Constraint relaxation is heuristic driven and based on the relational paradigm. Unlike RE IE, the indexing scheme TREE requires is not based on the membership of condition elements but on the sharing of references to symbols. On the basis of experimental evidence, TREE'S strategy showed better results than the standard RETE one. The number of comparisons during join steps has been reduced by a factor ranging from 1 to nearly two orders of magnitude. 1 Introduction Match continues to be a problem for AI systems, especially for general purpose production systems like SOAR [Laird et a/., 1987]. In this kind of systems, a large flexibility in representations is provided through their decomposition into binary relations or into an object-attribute-value paradigm. Consequently, basic data elements are mainly restricted to triples. As pointed out by Tambe and Rosenbloom [1990], this kind of restricted formalism leads to a combinatorial match since a condition element (CE) of a rule can match nearly all the working memory elements (WMEs), and condition parts of rules contain many CEs. Several works attempt to improve match algorithms for such systems. With or without paral-lelism, efficiency gains can be obtained through condi-TREE [Bouaud, 1992] is a state-saving match algorithm that has been designed for K, a production system with a restricted data formalism: WMEs are triples of symbols. TREE aims at reducing the size of join search spaces without using hashing techniques. It differs from RETE [Forgy, 1982] in its indexing scheme which does not rest on "memory support" 1 [McDermott et al., 1978], t.e. based on the CEs, but on the sharing of references to symbols. In this framework, we investigate a new join formulation. A constraint propagation and a heuristic relaxation based on the data paradigm is performed on the CEs in order to select the indexes to be used in the join step computation. These indexes are expected to be smaller than those built up from the CEs. The paper is organized as follows: section 2 …