Experience-based deductive learning

Abstract

A method of deductive learning is proposed to control deductive inference. The goal is to improve problem solving time by experience, when that experience monotonically adds knowledge to the knowledge base. Accumulating and exploiting experience are done by the schemes of knowledge migration and knowledge shadowing. Knowledge migration generates specific (migrated) rules from general (migrating) rules and accumulates deduction experience represented by specificity relationships between migrating and migrated rules. Knowledge shadowing recognizes rule redundancies during a deduction and prunes deduction branches activated from redundant rules. Three principles for knowledge shadowing are suggested, depending on the details of deduction experience representation.<>