INTEGRATING FEATURE-EXTRACTION AND MEMORY-SEARCH

Reasoning from prior cases or abstractions requires that a system iden tify relevant similarities between the current situation and objects r epresented in memory. Often, relevance depends upon abstract, thematic , costly-to-infer properties of the situation. Because of the cost of inference, a case-retrieval system needs to learn which descriptions a re worth inferring, and how costly the inference will be. This article outlines the properties that make an abstract thematic feature valuab le to a case-based reasoner, and recasts the problem of case retrieval into a framework under which a system can explicitly and dynamically reason about the cost of acquiring features relative to their informat ion value.