Method and apparatus for fast query approximation using adaptive query vector projection

摘要

Methods and apparatus are provided for querying by similarity in large collections of high-dimensional vectors by adaptively projecting each query vector into a library of projection elements; by selecting the set of elements that minimizes the work for processing each query; and by adaptively selecting the set of elements in order to exercise the tradeoff between query precision and query response time. The present invention provides for the storage of the projected vector as a meta-data description of data to enable content-based querying of the data.