Cost-aware query planning for similarity search

Similarity search aims to find all objects similar to a query object. Typically, some base similarity measures for the different properties of the objects are defined, and light-weight similarity indexes for these measures are built. A query plan specifies which similarity indexes to use with which similarity thresholds and how to combine the results. Previous work creates only a single, static query plan to be used by all queries. In contrast, our approach creates a new plan for each query. We introduce the novel problem of query planning for similarity search, i.e., selecting for each query the plan that maximizes completeness of the results with cost below a query-specific limit. By regarding the frequencies of attribute values we are able to better estimate plan completeness and cost, and thus to better distribute our similarity comparisons. Evaluation on a large real-world dataset shows that our approach significantly reduces cost variance and increases overall result completeness compared to static query plans.