Cardinality Estimation in Inner Product Space

This article addresses the problem of cardinality estimation in inner product spaces. Given a set high-dimensional vectors, query, and threshold, this estimates number vectors such that their products with query are not less than threshold. is an important for recent machine-learning applications maintain objects, as users items, by using matrices. The requirements solutions high efficiency accuracy. To satisfy these requirements, we propose sampling-based algorithm. We build trees via transformation to Euclidean space dimensionality reduction pre-processing phase. Then our algorithm samples existing nodes intersect search range on one trees. Our surprisingly simple, but it theoretically practically fast effective. conduct extensive experiments real datasets, results demonstrate shows superior performance compared techniques.