Optimized and Parallel Query Processing in Similarity-Based Databases

摘要

We present a novel method of query execution in similarity-based databases which adopts techniques commonly used in traditional programming language compilers. Our method is based on decomposition of relational algebra operators into a small set of simple operations which are subject of further optimizations. It shows up that with a small set of optimizations rules our system itself is able to infer efficient algorithms for data processing. Furthermore, operations we propose are compatible with the map/reduce approach to data processing, and thus, allows for implicitly parallel or distributed data processing.