Abstract
We explore the use of soft computing and user depned classipcations in multimedia database systems for content-based queries. With multimedia databases, due to subjectivity of human perception, an object may belong to different classes with different probabilities (ÒsoftÓ membership), as opposed to ÒhardÓ membership supported by conventional database systems. Therefore, we propose a uniped model that captures both hard and soft memberships. In practice, however, our model signipcantly increases the computation complexity (both online and off-line) and the storage complexity of content-based queries. Previously, we introduced a novel fuzzy-logic based aggregation technique to address the online computation complexity. In this paper, we propose novel techniques to cluster sparse user proples (i.e., items with missing data) to reduce both the off-line computation complexity and the storage complexity.