Cumulative learning techniques in production rules with fuzzy hierarchy (PRFH) system

Abstract

Cumulative learning, a promising route for automated knowledge acquisition and adaptation, involves using the results of prior learning to facilitate further learning. Achieving this objective would largely depend upon an enriched knowledge representation scheme. One of such efficient representations is Production Rules with Fuzzy Hierarchy (PRFHs) system. A PRFH, a standard production rule augmented with generality and specificity information, is of the form: where P is the set of preconditions = (P pub) k ∪ (P spl) k ∪ (P pvt) k and the specificity element Dki (di ) means that Dki is a specific class of Dk with degree of subsumption di. In this paper, a set of related PRFHs is called a cluster and is represented by a PRFH-tree. The proposed scheme incrementally incorporates new knowledge into set of clusters obtained from previous episodes and also maintains summary of clusters to be used in the future episodes. Using the Cumulative_growth algorithm, a new rule is added to the system, the Restructure_cluster algorithm restructures a cluster so as to minimize redundancy, and Merging_clusters algorithm enables merging of two related PRFHs clusters. The proposed system would be particularly useful in mining data streams and dynamic restructuring of knowledge bases.