Learning Flexible Concepts Using a Two-Tiered Representation

摘要: Most human concepts are flexible in the sense that they inherently lack precise boundaries, and these boundaries often context-dependent. This chapter describes a method for representing inductively learning from examples. The basic idea is to represent such using two-tiered representation. Such representation consists of two structures (“tiers”): Base Concept Representation (BCR), which captures explicitly context-independent concept properties, Inferential Interpretation (ICI), characterizes allowable modifications context-dependency. proposed has been implemented POSEIDON system (also called AQ16), tested on various practical problems, as “Acceptable union contracts” “Voting patterns Republicans Democrats U.S. Congress.” In experiments, generated descriptions were both, more accurate simpler than those produced by other methods tested, employing simple exemplar-based representations, decision tree learning, some previous rule learning.