Parallelization of Hierarchical Censored Production Rules

Abstract

A standard production rule is expressed in the form IF ltconditiongt THEN ltactiongt newlineProduction systems are widely used in Artificial Intelligence for modeling newlineintelligent behavior and building expert systems. However, standard production newlinesystems have a rigid structure as they cannot handle incomplete and imprecise newlineknowledge, which make them less flexible for adaptation. To capture the newlineuncertain and imprecise knowledge about the real world Michalski and Winston newlineintroduced the concept of Variable Precision Logic and suggested Censored newlineProduction Rules (CPRs) as an underlying representational and computational newlinemechanism to enable logic based systems to exhibit variable precision in which newlinecertainty varies while specificity remains constant. A CPR is a production rule newlineaugmented with exception conditions, with the following representation newlineIF ltconditiongt THEN ltactiongt UNLESS ltcensorgt newlinewhere ltcensofgt is an exception to the rule. newlineA CPR is quotunable to capture the taxonomic structure inherent in the knowledge newlineabout the real world. Bharadwaj and Jain have extended the concept of CPRs newlineby introducing two new operators to them, viz., GENERALITY and newlineSPECIFICITY to represent the more general and specific information, and newlinecalled them Hierarchical Censored Production Rules (HCPRs). HCPRs can be newlinemade to exhibit variable precision in reasoning such that both certainty in belief newlinein a conclusion and its specit1city may be controlled by the reasoning process. newlineThe general form of an HCPR is newlineIF B [ bl, b2, ... , bn] newlineTHEN A newlineUNLESS C [ cl, c2, ... ,en] newline{ preconditions } newline{decision I action} newline{ censor conditions } newlineii newlineGENERALITY G newlineSPECIFICITY S [ sl, s2, ... , sn] newline{ general information } newline{ specific information } newlineHCPR systems that support vanous symbolic and genetic based machine newlinelearning, have been found very useful in developing knowledge based systems, newlinewith learning capabilities and are capable of adjusting the certainty of inferences newlineto conform to time and other resource constraints. Such systems have newlinenumerous applications in situations