Combining Cache and Priority Queue to Enhance Evaluation of Similarity Search Queries

A variety of applications have been using content-based similarity search techniques. Higher effectiveness of the search can be, in some cases, achieved by submitting multiple similar queries. We propose new approximation techniques that are specially designed to enhance the trade-off between the effectiveness and the efficiency of multiple k-nearest-neighbors queries. They combine the probability of an indexed object to be a part of the precise query result and the time needed to examine the object. This enables us to improve processing times while maintaining the same query precision as compared to the traditional approximation technique without the proposed optimizations.