Data search method and related equipment

Abstract

The embodiment of the invention provides a data search method, which comprises the following steps of: constructing an index structure of a target vector table in a target database according to a preset index construction rule to obtain a target index structure of the target vector table; in response to a search request containing a to-be-queried vector, determining a current layer starting node in the topmost layer of the target index structure according to a preset method; aiming at the current layer, performing nearest neighbor search based on the starting node of the current layer to obtain a node closest to the vector to be queried in the current layer, and performing nearest neighbor search by taking the nearest node as the starting node of the current layer of the next layer; and when the current layer is the bottommost layer, executing greedy search by taking a node closest to the vector to be queried in all layers as an initial node to obtain a target search result of the query vector. The nearest neighbor search is carried out from the topmost layer to the lower layers in sequence, so that the search amount is reduced, and the vector search speed is increased.

Description

technical field [0001] The invention relates to the field of data search, in particular to a data search method and related equipment. Background technique [0002] With the development of informatization, more and more data needs to be stored in the database, which undoubtedly increases the computational cost of data search. In the field of image data analysis and processing, the scale of unstructured data is growing rapidly, such as video, image, voice, text, etc. With the rise of deep learning methods in various fields, most unstructured data can be converted into high-dimensional data through deep learning. The vectors are structured, and then through the vector search method, these vector data can be searched. In the current AIDB (Intelligent Database), the Top K search of high-dimensional vectors based on quantification is used, but the search time of the algorithm based on quantification will increase linearly with the growth of the data size, resulting in the failur...

AI Technical Summary

Problems solved by technology

In the current AIDB (Intelligent Database), the Top K search of high-dimensional vectors based on quantification is used, but the search time of the algorithm based on quantification will increase linearly with the growth of the data size, resulting in the failure of AIDB-based archiving services. The speed will also become slower and slower with the increase of data size. Therefore, in the case of high-speed growth of data size, the existing high-dimensional vector search has the problem of slow speed.

Images