A system, method and application for optimizing scalar-vector mixed queries on a vector database

By using an adaptive query optimization system, which leverages equal-depth histograms, Sketch, and deep learning models, the system optimizes scalar-vector hybrid queries in vector databases, solving the problems of unstable performance and high cost in existing technologies, and achieving more efficient query processing.

CN122285656APending Publication Date: 2026-06-26EAST CHINA NORMAL UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
EAST CHINA NORMAL UNIV
Filing Date
2024-12-25
Publication Date
2026-06-26

Smart Images

  • Figure HDA0005209690860000011
    Figure HDA0005209690860000011
  • Figure HDA0005209690860000012
    Figure HDA0005209690860000012
Patent Text Reader

Abstract

This invention discloses a scalar-vector hybrid query optimization system applied to a columnar-stored vector database. The system includes an optimizer, a cost estimator, an executor, and a analyzer. The optimizer rewrites and encodes the hybrid query, generating one or more query plans. It then calls the cost estimator to evaluate the cost of each query plan and selects the optimal plan after cost evaluation. The cost estimator evaluates the execution cost of each query plan and feeds back the evaluation result to the optimizer. The executor executes the selected optimal query plan and collects and records runtime information during execution. The analyzer analyzes the optimal query plan and the runtime information recorded by the executor to further optimize the cost estimator. This invention also discloses a scalar-vector hybrid query optimization method with wide application scenarios.
Need to check novelty before this filing date? Find Prior Art